0.1 adass2026 2026-11-01 2026-11-05 5 00:05 https://pretalx.adass.org/ https://pretalx.adass.org//media/adass2026/img/ADASS_XXXVI_EMU_LOGO_VARIATION_2_LIGHT_WUPMEVY_faLnWA6.webp Australia/Perth Seminar Room 1 Tutorial 2026-11-01T13:00:00+08:00 13:00 02:00 The tutorial will give a brief overview of the high-performance scientific programming language Julia. It will show how Julia solves the “two language problem” by compiling command-line expressions directly to machine code. It will touch on performing calculations using measurements (i.e., value-error pairs), units, symbolic manipulation, and GPU programming. It will conclude by showing how the combination of multiple dispatch and abstract types result in concise, generic, and high-performance software. The tutorial assumes a basic knowledge of scientific programming. adass2026-205-an-introduction-to-the-julia-programming-language Topical Computing, Software and Algorithms Paul Barrett en The objectives of the tutorial are to show how Julia is a versatile, general-purpose programming language and how function calls using multiple dispatch combined with abstract types, results in concise, modular, and highly performant code. false https://pretalx.adass.org/adass2026/talk/CKL3M9/ https://pretalx.adass.org/adass2026/talk/CKL3M9/feedback/ Banquet Hall Welcome 2026-11-02T09:00:00+08:00 09:00 00:05 Welcome from LOC, housekeeping and ADASS XXXVI kick-off. adass2026-195-opening-welcome-housekeeping-directions General Andreas Wicenec en Welcome from LOC, housekeeping and kick-off. false https://pretalx.adass.org/adass2026/talk/WZN9BT/ https://pretalx.adass.org/adass2026/talk/WZN9BT/feedback/ Banquet Hall Welcome 2026-11-02T09:05:00+08:00 09:05 00:15 Welcome to Country adass2026-196-welcome-to-country General Aaliyah Nestoridis en false https://pretalx.adass.org/adass2026/talk/KREDYT/ https://pretalx.adass.org/adass2026/talk/KREDYT/feedback/ Banquet Hall Welcome 2026-11-02T09:20:00+08:00 09:20 00:10 Welcome to UWA adass2026-197-welcome-to-uwa General Prof. Anna Novak, Deputy Vice-Chancellor Research UWA en false https://pretalx.adass.org/adass2026/talk/A9PFCR/ https://pretalx.adass.org/adass2026/talk/A9PFCR/feedback/ Banquet Hall Invited Talk 2026-11-02T09:30:00+08:00 09:30 00:30 In the 1980s there was an explosion in our ability to capture panoramic digital data from the sky and our ability to access cheap, powerful and flexible computers. We were awash in digital capabilities but held back by the lack of a similar explosive growth in square meters of glass and metal. This crisis changed the way ground based astronomy was done and led to astronomy leading the way in interoperability and scientific returns on digital assets. Now we may face another crisis as the cost of the data potential new telescopes can achieve, limits the science we can do. This will certainly change the way we preserve and value raw and science-ready data and the role of industry in advancing astronomy. I will review my journey in data intensive astronomy over the past 40 years and share some thoughts on future directions. adass2026-198-data-intensive-astronomy-data-flows-virtual-observatories-and-the-future-of-astronomical-data General Peter Quinn en false https://pretalx.adass.org/adass2026/talk/XRHS3X/ https://pretalx.adass.org/adass2026/talk/XRHS3X/feedback/ Banquet Hall Invited Talk 2026-11-02T10:45:00+08:00 10:45 00:30 Ray's talk adass2026-200-ray-s-talk The art of collaboration in astronomy software development Ray Brederode en false https://pretalx.adass.org/adass2026/talk/XVHVVN/ https://pretalx.adass.org/adass2026/talk/XVHVVN/feedback/ Banquet Hall Contributed Talk 2026-11-02T11:15:00+08:00 11:15 00:15 Radio astronomy in Australia is at a critical juncture. As SKA-Low nears completion, other key research instruments around the country are being significantly upgraded; there are now more components, procedures, and personnel involved in the production of our radio data products than ever before. Though significant in contribution, Australia’s radio astronomy industry has historically been small in size—driven by individuals with broad skills, deep institutional knowledge, and long tenures. As we enter this new era of scale, it is critical to understand the socio-technical factors that affect the radio interferometry data in Australia today, how they are likely to change as this next generation of personnel enters the field, and how this might affect the quality and applicability of the data they produce. Herein, we present findings from targeted interviewing of individuals who play a role in capturing and pre-processing radio astronomy data across Australia about their current workflows and planned changes—spanning those working in software development, data analysis, operations, and leadership across CSIRO, several universities, and SKA-related entities from SKAO to AusSRC. Looking at procedures for technical development, collaboration, prioritisation, validation, upskilling, knowledge transfer, and even succession planning, analysis showed healthy evolution in response to scale in some areas, while others seemed to diverge in ways that suggest a loss of lessons learned. We conclude with material recommendations from comparable fields of technical work, and by inviting audience members to participate in future phases of the project. adass2026-105-interrogating-processes-in-australian-radio-astronomy-data-production The art of collaboration in astronomy software development Mars Buttfield-Addison en false https://pretalx.adass.org/adass2026/talk/PMNWHP/ https://pretalx.adass.org/adass2026/talk/PMNWHP/feedback/ Banquet Hall Contributed Talk 2026-11-02T11:30:00+08:00 11:30 00:15 Astronomy has developed a rich ecosystem of specialised software for individual surveys, missions and scientific tasks. Combining these capabilities into a coherent researcher workflow, however, often requires **bespoke integration, duplication of functionality, or tight coupling** to another application's data structures and state. AstronomicAL was originally developed for interactive catalogue exploration, visualisation and human-in-the-loop active learning. Its evolution towards a more general **scientific workbench** created a new challenge: integrating independently developed scientific capabilities without turning the application into an increasingly coupled collection of tools. This became particularly concrete when combining Euclid-specific data access developed through ELSA with other independently developed analysis capabilities. We have therefore redesigned AstronomicAL as a **plugin host built around shared scientific contracts**. Extensions can retain their own implementation and dependencies while interacting through common abstractions for datasets and semantic mappings, object selection, derived scientific products and managed computation. Each plugin declares the scientific context it needs and the capabilities it contributes, while the platform mediates these interactions and keeps internal implementations independent. The current plugin set demonstrates this approach across community software and multiple surveys and missions. Aladin Lite can follow the selected source through generic sky-coordinate mappings; catalogue data can be exchanged with TOPCAT through SAMP; common spectral tooling accesses DESI, SDSS and Euclid data; and Euclid cutouts can be used alongside generic visualisation, annotation and machine learning capabilities. **Active learning, previously central to AstronomicAL, is now one composable scientific workflow over the same platform.** We discuss the interfaces and boundaries required for independent tools to work together: which scientific concepts need to be shared, which implementation details should remain independently owned, and how plugin-host architectures can enable astronomy tools to be composed into broader scientific workflows. adass2026-167-astronomical-a-plugin-based-workbench-for-composable-astronomy-workflows The art of collaboration in astronomy software development /media/adass2026/submissions/LALT7Q/AstronomicAL_ADASS_jLQjtEf_uBCBsQA.webp Grant Stevens en false https://pretalx.adass.org/adass2026/talk/LALT7Q/ https://pretalx.adass.org/adass2026/talk/LALT7Q/feedback/ Banquet Hall Contributed Talk 2026-11-02T11:45:00+08:00 11:45 00:15 The Square Kilometre Array Observatory (SKAO) is a geographically distributed scientific programme involving hundreds of scientists, systems engineers, software developers, operators, and stakeholders across multiple countries and organisations. Developing software at this scale presents not only technical challenges, but also challenges in building and maintaining a shared understanding of systems, priorities, risks, and responsibilities across diverse communities. Over multiple Planning Increments (PIs), we found that conventional artefacts such as requirements documents, Jira tickets, planning boards, architectural diagrams, and status reports were often insufficient to communicate the relationships and dependencies inherent in a complex observatory software ecosystem. Different stakeholders frequently possessed different mental models of the same problem, leading to misalignment despite access to the same information. These visual artefacts were applied throughout the software lifecycle -from onboarding new team members and communicating software architecture to Programme Increment planning, capacity allocation, dependency management, release coordination, subsystem interaction modelling, operational architecture, non-functional requirements, and clarifying roles and responsibilities. Each visual was created to simplify a specific communication challenge by making complex relationships easier to understand and discuss. The work presented is not intended as an alternative to existing software engineering practices or tools, but rather as a complementary collaboration technique for large scientific software projects. As astronomy software ecosystems continue to grow in complexity and scale, visual thinking offers a lightweight yet effective way to help distributed communities answer a fundamental question: Are we all looking at the same sky? adass2026-81-seeing-the-same-sky-visual-thinking-as-a-collaboration-tool-in-the-skao The art of collaboration in astronomy software development /media/adass2026/submissions/F8FNGH/ska-science-software-diagram_SHaaB_OkAhjB9.webp Snehal Valame en false https://pretalx.adass.org/adass2026/talk/F8FNGH/ https://pretalx.adass.org/adass2026/talk/F8FNGH/feedback/ Banquet Hall Invited Talk 2026-11-02T13:30:00+08:00 13:30 00:30 Astronomical literature is expanding at an unprecedented rate, with thousands of papers added every month to preprint servers like arXiv.org and indexed by the NASA Astrophysics Data System (ADS). For academics and students, staying current with relevant work while keeping track of shifting trends therefore represents a critical challenge. I will talk about lessons learned from developing Pathfinder, a complement to systems like ADS that uses large language models combined with retrieval-augmented generation (RAG) to enable semantic search and question-answering across the astronomy literature. I will discuss some of the unique challenges of applying NLP and LLMs to scientific publications in astronomy, including grounding LLM responses in published literature to minimize hallucinations, and leveraging embeddings to create interpretable semantic spaces for literature exploration. Drawing from Pathfinder's deployment (pfdr.app) and user feedback from the astronomy community, I will highlight how interpretable intermediate representations such as semantic embeddings and citation graphs can lend interpretability and rigor to otherwise black-box models, and help their adoption in research pipelines. adass2026-201-using-insights-from-the-embedding-spaces-of-large-language-models-for-astronomical-research-and-discovery AI as a tool for data discovery and data management Kartheik Iyer en false https://pretalx.adass.org/adass2026/talk/3JPSP9/ https://pretalx.adass.org/adass2026/talk/3JPSP9/feedback/ Banquet Hall Contributed Talk 2026-11-02T14:00:00+08:00 14:00 00:15 The rapid adoption of generative AI is transforming the way scientists interact with astronomy science archives and platforms. However, the growing dependence on large foundation models raises important questions about sustainability, cost, and long-term operational viability. In the Data Science and Archives Division at the European Space Astronomy Centre (ESAC), we are exploring different ways to use AI to support our users. Among others, we are exploring how smaller open source language models combined with retrieval-augmented generation (RAG) and domain-specific tooling can enhance astronomical data discovery while minimizing computational overhead. Rather than treating state of the art frontier models as the default solution, we are investigating if smaller models can be effectively deployed for many archive-support tasks with considerably lower infrastructure requirements. The ESAC astronomy science archives present a particularly attractive use case for this approach. In legacy missions, the number of specialists available to support users decreases with time, while the scientific value of these missions remains high for decades. By combining lightweight large language models (LLM) with RAG pipelines built from mission documentation, archive interfaces user guides, and selected scientific publications, it is possible to provide conversational interfaces that preserve mission knowledge that can be used by scientists to discover and use relevant datasets. Similarly, for missions in operations, such as Euclid, AI assistants could help scientists navigate its complex data model, learn the details of hundreds of distinct data products, and generate ADQL example queries and notebooks. We argue that this approach could complement other alternatives and represents a more sustainable path for scientific archives. Running open-source LLMs with “only” a few billion parameters on local GPU infrastructure would allow institutions to reuse existing resources, reduce dependence on commercial AI subscriptions, and limit the environmental and financial costs associated with large-scale cloud inference. This approach also maximizes the value of open-source models whose development has already required significant community investment. In this presentation we will showcase some of the experimental work we are doing at ESAC using open source models running on our local infrastructure, for example, enhancing the existing ESASky chatbot with generative AI capabilities and building RAG pipelines for mission documentation and archival data discovery. adass2026-134-sustainable-ai-for-astronomical-archival-data-discovery-in-the-petabyte-era AI as a tool for data discovery and data management Marcos López-Caniego en false https://pretalx.adass.org/adass2026/talk/QG8EJW/ https://pretalx.adass.org/adass2026/talk/QG8EJW/feedback/ Banquet Hall Contributed Talk 2026-11-02T14:15:00+08:00 14:15 00:15 The NASA Exoplanet Archive (NEA) is developing a suite of tools to speed up and streamline the process of getting data from published papers into ingestion-ready form. Manually locating relevant papers, extracting tabular and textual data, and reformatting it to archive standards is a major bottleneck in keeping the archive current. We describe three complementary efforts underway to address this in different parts of our pipeline, two of which utilize the capabilities of modern LLMs. First, we have created a standardized template that authors can use to submit their published data directly in our required ingestion format, reducing the translation burden on archive staff and ensuring parameter accuracy. Second, we have built an AI tool that scans the weekly astronomical literature to identify papers most likely to contain data of interest for NEA, cutting down on manual literature triage. Third, we are developing an AI tool that parses the text and tabular content of identified papers and converts it into the structured format required for ingestion. Fourth, we are writing tools to facilitate bulk ingest of large sets of curated data, such as GAIA DR4. Together, these tools aim to shorten the pipeline from publication to archive availability, reduce manual effort, and improve the completeness and timeliness of the NEA holdings. We will present the basic design of each tool, early results, and lessons learned in applying AI to a production data-curation workflow. adass2026-146-new-tools-for-streamlining-data-ingestion-at-nasa-exoplanet-archive-nea AI as a tool for data discovery and data management Meca Lynn en false https://pretalx.adass.org/adass2026/talk/ETDUGR/ https://pretalx.adass.org/adass2026/talk/ETDUGR/feedback/ Banquet Hall Contributed Talk 2026-11-02T14:30:00+08:00 14:30 00:15 Astronomical data discovery often requires identifying appropriate archives, resolving astronomical object names and coordinates, translating observational constraints into service-specific query parameters, and interpreting complex metadata returned by heterogeneous data services. Although the Astroquery Python package provides programmatic access to many astronomical archives, coordinating these resources still requires detailed knowledge of individual interfaces. We present AstroAKD, a data-search agent developed within NASA IMPACT’s Accelerated Knowledge Discovery (AKD) ecosystem, a modular platform for developing, evaluating, and deploying scientific AI agents that connect natural-language reasoning with external data and software services. AstroAKD employs a [Model Context Protocol (MCP) server](https://github.com/nasa-impact/astroquery-mcp) built on Astroquery that dynamically exposes supported archive functions to a large language model (LLM) agent. At runtime, the agent discovers available functions, inspects their parameters, converts structured requests into astronomy-specific objects such as coordinates and unit-bearing quantities, and normalizes returned tables and metadata into machine-readable responses. Supported services include SIMBAD, ADS, MAST, HEASARC, IRSA, NED, Gaia, VizieR, and Virtual Observatory resources. Given a scientific request, AstroAKD resolves target identifiers, extracts observational constraints, selects appropriate services, constructs executable queries, and returns relevant metadata and candidate datasets with archive provenance. We describe the agent architecture and evaluate representative single- and multi-archive workflows using archive and function selection, parameter correctness, task completion, and recovery from ambiguous identifiers, empty results, and unavailable services. Benchmark requests, tool calls, and outputs are retained as inspectable traces for validation and repeatability. AstroAKD demonstrates how domain-specific AI agents and standardized tool interfaces can support transparent, rapid exploration of astronomical archives. adass2026-168-astroakd-an-mcp-based-agent-for-natural-language-search-of-astronomical-archives AI as a tool for data discovery and data management Ashkbiz Danehkar en false https://pretalx.adass.org/adass2026/talk/WR3GKK/ https://pretalx.adass.org/adass2026/talk/WR3GKK/feedback/ Banquet Hall Contributed Talk 2026-11-02T14:45:00+08:00 14:45 00:15 As implementing complex AI models with massive datasets becomes increasingly accessible, at what point do we reach the limits of explainability? "Simple" problems are easily tackled by AI when input parameters are well-defined and the underlying processes are understood. However, as datasets grow and the underlying processes become exponentially ill-defined, we must question whether we are too quick to abandon explainability in favor of rapid results. While societal pressures often encourage this trade-off, the scientific community must resist it. Furthermore, our human drive to seek out rules that lead to definitive conclusions can be misleading. Studies have shown that rats can easily outperform humans at certain pattern recognition tasks, illustrating how our own cognitive biases can interfere with objective analysis. This presentation will explore these challenges using examples of analysing over 100 million astronomical sources via multi-modal AI models, demonstrating how an abundance of information can sometimes yield increasingly misleading results. I will also discuss how we utilise AI for our SRCnet software and data management stack at SKAO, specifically examining the consequences of agentic coding and how solving one problem inevitably introduces unintentional side effects. Ultimately, from analysing data to developing software, mitigating these unintentional consequences and inherent biases has become the primary focus and responsibility of human problem-solving in the era of AI. adass2026-158-unexplainable-ai-rats-super-recognisers-and-multi-modal-deep-learning-for-astronomical-datasets AI as a tool for data discovery and data management Alex Clarke en false https://pretalx.adass.org/adass2026/talk/97NUVC/ https://pretalx.adass.org/adass2026/talk/97NUVC/feedback/ Banquet Hall Contributed Talk 2026-11-02T16:00:00+08:00 16:00 00:15 Radio interferometric pipelines encode expert judgment as fixed heuristics: reasonable defaults for a "typical" dataset. When a dataset doesn't fit the mold, an expert inspects the diagnostics and steps in by hand. That doesn't scale as surveys push toward higher data volumes and less human oversight. LLMs can close that gap. They read the same diagnostics an expert would and reason about them per dataset. But letting a model both reason and act is risky. LLMs hallucinate, drift, and skip steps silently. We built an architecture that strictly separates measurement from reasoning. The model gets room to adapt, but it never touches the data path directly. An MCP layer wraps the reduction software (CASA, here) and exposes each Measurement Set operation (metadata queries, instrument geometry, calibration, imaging) as an independent tool. Every tool returns structured data with explicit completeness and provenance. None of them interpret their own output or call each other. Reasoning lives outside the tools, in skills: version-controlled, plain-text documents that encode interferometric expertise. They're fed into the LLM's context stage by stage, so it can reason about what the tools hand back. An orchestration layer walks the model through a sequence of stages that looks like a processing pipeline, except the parameters aren't fixed. Skill-based reasoning lets the model make an informed, per-dataset call instead of falling back on defaults. Every stage's state, its parameter choices, and the sequencing are written out as documentation and a reproducible script. Since reasoning lives outside the MCP layer, the orchestrator doesn't care which model is driving it. Cloud (Claude, Codex) or local and open (Gemma, Qwen): both work. A model equipped with real domain expertise, via skills, beats a one-size-fits-all pipeline, dataset by dataset. We demonstrate end-to-end inspection and calibration on VLA, GMRT, and ALMA data. We discuss extending the approach to other instruments, facilities, and pipelines. adass2026-154-llm-orchestrated-radio-interferometric-data-reduction AI as a tool for data discovery and data management /media/adass2026/submissions/ZS9EYM/G55_radio_analyst_8qNPxuZ_nfGGYqn.webp Srikrishna Sekhar en The GitHub repository can be found here - https://github.com/skunkworks-ra/radio-analyst false https://pretalx.adass.org/adass2026/talk/ZS9EYM/ https://pretalx.adass.org/adass2026/talk/ZS9EYM/feedback/ Banquet Hall Contributed Talk 2026-11-02T16:15:00+08:00 16:15 00:15 Generating database queries from natural language remains a challenging task, particularly in specialized scientific domains. In this work, we study natural language to Astronomical Data Query Language (ADQL) generation using large language models (LLMs). We curate a high-quality dataset of natural language–ADQL pairs through an LLM-assisted filtering and validation pipeline and use it to fine-tune models of varying sizes and capabilities. To enable systematic evaluation, we construct an expert-annotated benchmark of queries for the Gaia mission, spanning a range of query complexities, from simple retrieval tasks to complex joins and aggregations. Finally, we compare fine-tuned models against retrieval-augmented generation (RAG) approaches, analyzing their effectiveness in terms of query correctness and robustness. Our results provide insights into the relative strengths of fine-tuning and retrieval augmentation for domain-specific scientific query generation. adass2026-175-adql-generation-using-llms AI as a tool for data discovery and data management Sandor Kruk en false https://pretalx.adass.org/adass2026/talk/WUYEGC/ https://pretalx.adass.org/adass2026/talk/WUYEGC/feedback/ Banquet Hall Contributed Talk 2026-11-02T16:30:00+08:00 16:30 00:15 The ESA Near-Earth Object Coordination Centre (NEOCC) is responsible for the monitoring of Near-Earth Objects and the assessment of potential impact hazards. These activities rely on Aegis [1], a software system for asteroid orbit determination, ephemerides generation, uncertainty propagation, close-approach analysis, and impact monitoring. The results and computational capabilities of Aegis are made available to the community through the NEOCC web portal [2] and dedicated APIs [3]. As operational demands evolved, Aegis transitioned from a standalone scientific application to a service deployed within ESA's cloud infrastructure. This contribution presents the architectural evolution of Aegis and the engineering decisions that enabled its migration to a distributed operational environment while preserving a mature and validated scientific code base. The original system was conceived as a largely monolithic application running on a single machine. Rather than pursuing a complete redesign, the development strategy focused on incremental modernisation, gradually introducing containerisation, automated testing and deployment workflows, service-oriented interfaces, and distributed execution capabilities. Today, Aegis operates as a collection of Docker-based services deployed on ESA Cloud using Docker Swarm orchestration. Scientific processing remains largely performed by established Fortran components, while Python services coordinate workflows, expose REST interfaces, and exchange information through Redis-based messaging queues. Although parts of the system still rely on shared file-system access, reflecting historical design choices, this approach has proven adequate for current operational requirements and data volumes while allowing progressive migration towards more centralised data management solutions. The software has also been deployed on ESA High-Performance Computing (HPC) infrastructures, requiring targeted adaptations but without fundamental changes to its scientific components. Modern software engineering often favours highly scalable cloud-native solutions. However, scientific operational systems frequently evolve under different constraints, including long-lived code bases, limited development resources, strict reliability requirements, and moderate data volumes. The Aegis experience shows that significant gains in maintainability, deployment flexibility, and operational robustness can be achieved through incremental modernisation, without the need for unnecessarily complex architectures. [1] https://doi.org/10.1007/s10569-024-10225-z [2] https://neo.ssa.esa.int/ [3] https://neo.ssa.esa.int/computer-access adass2026-120-engineering-for-requirements-the-evolution-of-the-aegis-asteroid-impact-monitoring-system-at-esa Topical Computing, Software and Algorithms Francesco Gianotto en false https://pretalx.adass.org/adass2026/talk/MUDUE9/ https://pretalx.adass.org/adass2026/talk/MUDUE9/feedback/ Banquet Hall Contributed Talk 2026-11-02T16:45:00+08:00 16:45 00:15 We present a new GPU-accelerated signal processing pipeline for real-time beamforming, correlation, and radio frequency interference (RFI) mitigation through spatial filtering on densely packed aperture arrays and phased array feeds. Developed as an instrument-agnostic pipeline, we demonstrate the pipeline's application to the Low-frequency Australian Megametre-Baseline Demonstrator Array (LAMBDA), a new 50–350 MHz VLBI-capable array presently commissioning 36 antennas at the Paul Wild Observatory, Narrabri, New South Wales, Australia. While the theory of spatial-filtering-based RFI mitigation is well established, building a working real-time system raises genuine systems engineering questions: how to structure a pipeline flexible enough to support rapid iteration on detection and mitigation strategies; how to cleanly inject a priori information, such as sky models or known interferer locations, into a high-performance pipeline; and how to balance flexibility against the throughput demanded by real-time operation. LAMBDA's current modest scale and short baselines make it an ideal testbed for iterating on these implementation problems before deployment at larger arrays. We discuss the design of the pipeline and the trade-offs involved in making it both flexible and fast. We also benchmark its GPU throughput, showing performance sufficient for processing 25 MHz of bandwidth on one NVIDIA A100 for both the current and full 256-antenna LAMBDA array. Using LAMBDA data, we demonstrate the pipeline's RFI mitigation on real interfering signals. adass2026-132-spatial-filtering-in-practice-a-gpu-pipeline-for-real-time-rfi-mitigation Topical Computing, Software and Algorithms Jay Smallwood en false https://pretalx.adass.org/adass2026/talk/AQYPKM/ https://pretalx.adass.org/adass2026/talk/AQYPKM/feedback/ Poster Rooms Poster Session 2026-11-02T10:00:00+08:00 10:00 00:45 1st poster session adass2026-199-poster-session-1 Poster en false https://pretalx.adass.org/adass2026/talk/HPWPYC/ https://pretalx.adass.org/adass2026/talk/HPWPYC/feedback/ Poster Rooms Poster Session 2026-11-02T15:00:00+08:00 15:00 01:00 Poster Session 2 adass2026-202-poster-session-2 Poster Andreas Wicenec en false https://pretalx.adass.org/adass2026/talk/YSMMTC/ https://pretalx.adass.org/adass2026/talk/YSMMTC/feedback/ Banquet Hall Invited Talk 2026-11-03T09:00:00+08:00 09:00 00:30 Placeholder adass2026-213-christoph-s-talk AI as tool for software engineering Christof Buchbender en false https://pretalx.adass.org/adass2026/talk/FBQB9X/ https://pretalx.adass.org/adass2026/talk/FBQB9X/feedback/ Banquet Hall Contributed Talk 2026-11-03T09:30:00+08:00 09:30 00:15 In [ADASS 2019, Shortridge](https://www.aspbooks.org/a/volumes/article_details?paper_id=39885) benchmarked twelve programming languages on a trivial 2-D-array kernel and *found a million-fold range of execution speeds* - the sort of result that quietly reshapes how one thinks about astronomy code performance. We have reprised that study seven years later, on modern Apple-Silicon hardware, *expanded it to twenty-nine languages*, and - as a first for ADASS - used a large language model as porting and analysis partner. The most useful new finding for working astronomers is that the choice that matters most is no longer which language, but which runtime. A Python loop can now run two or three orders of magnitude faster than the unmodified CPython baseline simply by swapping the interpreter or adding a JIT decorator - no rewrite, no C extension, no language switch. We discuss what this implies for scientific Python practice, what the AI co-author contributed, and what it did not. adass2026-72-same-source-three-orders-of-magnitude-a-2026-reprise-of-shortridge-s-compiler-benchmark-co-authored-with-an-llm AI as tool for software engineering Karl Glazebrook en false https://pretalx.adass.org/adass2026/talk/UZPMJZ/ https://pretalx.adass.org/adass2026/talk/UZPMJZ/feedback/ Banquet Hall Contributed Talk 2026-11-03T09:45:00+08:00 09:45 00:15 Building and optimizing High-Performance Computing (HPC) pipelines for next-generation radio telescopes (e.g., SKA) presents immense software engineering challenges. We present a novel approach using continuous "Loop Engineering" driven by specialized Large Language Model (LLM) agents to refactor a radio interferometry imaging codebase (RICK). By deploying distinct AI personas—such as an HPC/MPI specialist, a GPU-porting engineer, and a domain-specific radio astronomer—we automated complex architectural upgrades. Operating sequentially to prevent code conflicts, these agents successfully aligned scientific outputs with industry standards (WSClean) within a strict 1% tolerance, guaranteed MPI scale invariance, and optimized OpenMP GPU memory traffic. We discuss the efficacy of agent-driven automated development, the technique of prompt cross-validation across different LLM engines to debug complex FITS WCS metadata, and the paradigm shift from manual coding to managing autonomous AI engineering loops in astronomical software development. adass2026-82-multi-agent-loop-engineering-for-radio-astronomy-pipelines-autonomous-refactoring-and-scientific-validation AI as tool for software engineering /media/adass2026/submissions/9ELWAF/ADASS_Agentic_AI_Radio_rpejtrU_1KC_Kgs4zP0.webp Giovanni Lacopo en The pipeline used as a benchmark for this experiment (RICK) is a highly modular C++ computational engine designed for high-resolution imaging and deconvolution of massive radio astronomy datasets. The autonomous multi-agent framework allowed an engineer transitioning from traditional HPC to AI-driven workflows to address both low-level parallel optimizations (MPI/GPU) and high-level astrophysics validation (such as WCS coordinate systems and flux calibration alignment) within the same development cycle. Crucially, the AI infrastructure was also leveraged to explore future hardware horizons, evaluating how the code's modular structure could be prepared for the upcoming computational paradigm shifts. false https://pretalx.adass.org/adass2026/talk/9ELWAF/ https://pretalx.adass.org/adass2026/talk/9ELWAF/feedback/ Banquet Hall Contributed Talk 2026-11-03T11:00:00+08:00 11:00 00:15 In 2026 large language model coding agents such as Codex and Claude have developed rapidly and changed the face of software development. They can be used for debugging, performance optimization, and adding features to existing codebases. As the tooling improved we decided to see if we could use these systems to port a large C library to Rust. The library we chose was the Starlink AST library, a library that consists of 200,000 lines of C (not including comments or blank lines) that has been in development for over 30 years and contains some very specific fixes for very rare combinations of FITS headers. In this talk I will tell the story of agents getting confused (but also agents improving), agents cutting corners and hoping I wouldn't notice, agents finding bugs in the original code but then making sure the Rust had the same bugs, eye-watering token costs, and how the real lesson from this journey is that your ported code is only as good as the test suite of your original library. adass2026-97-using-llms-to-port-a-large-c-library-to-rust AI as tool for software engineering Tim Jenness en false https://pretalx.adass.org/adass2026/talk/WZLDXB/ https://pretalx.adass.org/adass2026/talk/WZLDXB/feedback/ Banquet Hall Invited Talk 2026-11-03T11:15:00+08:00 11:15 00:30 Rubin is now about one year in operations and DM has delivered several Data Previews including DP2 which contains the extensive and good commissioning data. We will show some results (yes pretty pictures ) of our superb pipelines, discuss some of the challenges of moving and processing TBs of data from Chile and to serving Ks of users. We will perhaps mention a few lessons learned. Rubin has also transitioned from a stand alone project to being embedded in an organisation - we shall ponder organisational inertia and team vs organisational clashes. adass2026-208-rubin-data-management-delivering-awe-and-wonder-and-weathering-organisational-upheaval Global data management and lifecycle in the exascale era /media/adass2026/submissions/3X87MG/54538837528_64cf835895_c_bXIjS3N_SjrpMiT.webp William O Mullane en I will do a google deck I intend to include some videos .. in theory it will rin on any machine but may need a checkout. false https://pretalx.adass.org/adass2026/talk/3X87MG/ https://pretalx.adass.org/adass2026/talk/3X87MG/feedback/ Banquet Hall Contributed Talk 2026-11-03T11:45:00+08:00 11:45 00:15 Astronomers historically, and sometimes for good reasons, tend to keep every bit of data, forever! In some cases to the degree that you'll find a cupboard full of 9-track tapes or even punch-cards, or a locked-down vault full of deteriorating photographic plates with hand-written observer log-books. Even if there would be valuable data on there, it would be extremely hard to get to it, or even just knowing about its existence. Any form of data needs active data management and stewardship to make and keep it findable and accessible. For some of the latest facilities the on-going costs of data management has become a significant fraction of the overall operational costs. On the other hand the value of any data changes over time. Think about yesterday's newspaper. Acknowledging this, the SKA is implementing a Data Lifecycle Management system (DLM) rather than a Data Management system. On the technical side this talk is about that implementation but with a focus on controversial topics like lossy compression and data deletion, informed by a defined life-cycle for every data item in the system. adass2026-106-astronomers-unfortunately-you-need-to-change-your-hoarder-mentality Global data management and lifecycle in the exascale era Andreas Wicenec en false https://pretalx.adass.org/adass2026/talk/7PXYSY/ https://pretalx.adass.org/adass2026/talk/7PXYSY/feedback/ Banquet Hall Contributed Talk 2026-11-03T13:30:00+08:00 13:30 00:15 The next generation of radio astronomy arrays are challenging existing data analysis paradigms, as they have an order of magnitude more antennas and larger bandwidth. For these instruments, data storage will be the major cost driver and will constrain the processing. The traditional imaging methods for deep (multi-epoch) observations require more storage than is available for ASKAP and than will be available for SKA. On the other hand reducing the data volume by producing an image cube from every observing epoch bakes in systematic errors and (if the errors are non-Gaussian) imposes a sensitivity limit that can be significantly higher than the science requirement. For this reason the radio astronomers are demanding that they retain access to the visibilities for processing the data. The only way that this can be affordable is if the data volumes for visibilities are reduced to a level that is manageable. We have been testing two methods of data compression, which will dramatically reduce the storage requirements: grid-stacking, a two-stage lossless compression solution; and lossy compression of the raw visibilities, before traditional processing. Both are providing excellent results. The lossless grid stacking data product, after compression, are twenty times smaller than the individual measurement sets. The introduced losses in the recovered HI spectra are about 1\%. We are using this for DINGO. The lossy compression can provide measurement-sets that are ten times smaller than the individual measurement sets and the losses introduced are much less than 0.1\%. We are using this on MeerKAT. adass2026-136-a-little-can-go-a-long-way-investigations-into-visibility-compression-for-radio-interferometry-data-storage Global data management and lifecycle in the exascale era Richard Dodson en false https://pretalx.adass.org/adass2026/talk/QRNT8H/ https://pretalx.adass.org/adass2026/talk/QRNT8H/feedback/ Banquet Hall Invited Talk 2026-11-03T13:45:00+08:00 13:45 00:30 Placeholder adass2026-210-omkar-s-talk AI as tool for scientific discovery Omkar Bait en false https://pretalx.adass.org/adass2026/talk/HYLWYE/ https://pretalx.adass.org/adass2026/talk/HYLWYE/feedback/ Banquet Hall Contributed Talk 2026-11-03T14:15:00+08:00 14:15 00:15 Radio interferometric imaging is an ill-posed inverse problem because the sky brightness distribution is sampled sparsely and irregularly in the Fourier (uv) domain. Incomplete uv-coverage contributes to imaging artefacts, uncertain flux recovery, and distortions of source morphology, particularly for faint and extended emission. Many machine-learning approaches operate on dirty or reconstructed images, after instrumental effects and information loss have already been incorporated into the image representation. Learning directly from complex visibilities offers an alternative: it allows the model to represent the measurement domain continuously and to infer unsampled Fourier components under a learned prior before conventional image formation. We present the adaptation and high-performance-computing evaluation of a transformer-conditioned neural field for continuous visibility-domain modeling. Sparse complex visibilities, together with their uv coordinates, are encoded by a Transformer into a latent representation. This representation conditions a multilayer perceptron through feature-wise linear modulation. The resulting neural field can be queried at arbitrary Fourier coordinates and therefore does not restrict the reconstructed visibility function to a fixed grid. We evaluate the approach using simulated observations designed to reproduce key characteristics of LOFAR HBA data. The network is implemented and tested on a multi-GPU HPC system. Our study examines model fidelity and generalization to unsampled coordinates and the computational limits imposed by the number of input visibilities and target image resolution. We also investigate strategies for compressing large visibility datasets into tractable latent representations while retaining information relevant across multiple spatial scales. We will present preliminary visibility-modeling results, computational and memory-scaling benchmarks, and the methodological developments required to extend the approach to real interferometric observations. These results identify both the potential of continuous visibility representations and the remaining challenges associated with data volume and realistic instrumental sampling. adass2026-124-toward-scalable-neural-fields-for-continuous-visibility-modeling-in-radio-interferometry AI as tool for scientific discovery Nicoletta Sanvitale en false https://pretalx.adass.org/adass2026/talk/X93GY8/ https://pretalx.adass.org/adass2026/talk/X93GY8/feedback/ Banquet Hall Contributed Talk 2026-11-03T14:30:00+08:00 14:30 00:15 Radio interferometers and wide-area continuum surveys are producing increasingly large and complex imaging datasets, making manual source identification and classification impractical. Accurate, fast and fully automated methods are therefore required to detect and characterise two distinct but related populations: extended radio sources associated primarily with active galactic nuclei, including jets and radio lobes, and diffuse low-surface-brightness sources, such as radio halos, relics, bridges and filamentary emission associated with galaxy clusters and the cosmic web. Both classes exhibit complex, irregular and frequently multi-component morphologies that are difficult to recover with conventional source-finding algorithms. TUNA (TransUNet for Astrophysical data) has been introduced as a deep-learning framework for the automated segmentation of radio sources. Its effectiveness has already been investigated across all the targeted source classes, including extended radio galaxies and diffuse halos, relics and bridges. TUNA is based on a hybrid TransUNet architecture combining a convolutional encoder, a Vision Transformer and a U-Net-like decoder. Convolutional layers extract local brightness and morphological features, while multi-head self-attention captures long-range spatial dependencies between image regions. Skip connections preserve fine spatial information during the reconstruction of full-resolution pixel-level segmentation maps. The TUNA training and inference pipelines are designed to exploit GPU acceleration, multi-GPU data parallelism and mixed-precision computation. Computationally intensive convolutional and attention operations are executed on GPUs, while the workload can be distributed across multiple devices. Mixed-precision execution effects on memory and power consumption, and computational performance is being investigated. This contribution will present the application of TUNA to the systematic processing of LOFAR Two-metre Sky Survey Data Release 3 pointings. The talk will describe the processing workflow and assess the method from both scientific and computational perspectives. Its effectiveness in detecting and segmenting extended and diffuse radio sources will be evaluated together with its computational efficiency, scalability, GPU utilisation and multi-GPU performance. Scientific accuracy and computational performance will be presented and discussed as complementary requirements for the deployment of AI-based source-segmentation methods in current and future large-area radio surveys. adass2026-125-scalable-deep-learning-segmentation-of-extended-and-diffuse-radio-sources-with-tuna AI as tool for scientific discovery Claudio Gheller en false https://pretalx.adass.org/adass2026/talk/MWJB3T/ https://pretalx.adass.org/adass2026/talk/MWJB3T/feedback/ Banquet Hall Contributed Talk 2026-11-03T14:45:00+08:00 14:45 00:15 Modern radio continuum surveys such as ASKAP's Evolutionary Map of the Universe (EMU) produce catalogues of tens of millions of sources, far exceeding what manually curated pipelines can process. In this talk, I will present two complementary systems addressing this. RG-CAT uses an object-detection machine-learning model as its backbone to detect and classify radio galaxies — including complex, extended morphologies — directly from survey imaging, feeding into cataloguing scripts that produce science-ready outputs. EMUSE (the Evolutionary Map of the Universe Search Engine) uses a multimodal model to generate and search through image embeddings, making survey catalogues discoverable by the wider research community. These are deployed as containerised, cloud-native services, illustrating how machine learning can serve as the backbone of scalable, end-to-end research pipelines. adass2026-164-from-detection-to-discovery-ml-pipelines-for-the-radio-sky-surveys AI as tool for scientific discovery Nikhel Gupta en false https://pretalx.adass.org/adass2026/talk/RHQVJB/ https://pretalx.adass.org/adass2026/talk/RHQVJB/feedback/ Banquet Hall Contributed Talk 2026-11-03T16:15:00+08:00 16:15 00:15 Radio interferometric imaging reconstructs sky brightness from sparsely sampled Fourier measurements, leading to a highly ill-posed inverse problem that is increasingly challenged by the scale and resolution demands of next-generation telescopes, such as the Square Kilometre Array (SKA). Classical approaches, including CLEAN and its variants, often struggle with extended emission and face significant challenges in scaling to high-throughput, high-dynamic-range observations. We propose a data-driven transformer-based radio interferometric imaging algorithm, TransformerRIM, that integrates learnt image priors with physics-based measurement constraints. The reconstruction module is built on a Swin Transformer encoder-decoder, enabling multi-scale feature extraction and long-range spatial modelling. We further implement CUDA-based differentiable forward and adjoint operators for mapping between image space and irregularly sampled visibilities, allowing residual visibility information to be incorporated into recurrent reconstruction and end-to-end training. Uncertainty quantification is performed during inference by sampling perturbed visibilities from noise models, reconstructing a set of possible images. The method is trained on simulated and real datasets from JVLA B-configuration and LOFAR. On data of 3C 75 (a binary SMBH system), the inference result can achieve a dynamic range of 5.91M using our model, compared to 3.01M using CLEAN. The trained model makes inference substantially faster than CLEAN. The experimental results demonstrate that TransformerRIM provides a scalable and reliable path toward high-efficiency, high-dynamic-range radio imaging pipelines for next-generation radio telescopes. adass2026-93-transformerrim-a-data-driven-transformer-based-radio-interferometric-imager-with-uncertainty-quantification AI as tool for scientific discovery Qitong Anabel Tan en false https://pretalx.adass.org/adass2026/talk/XRSQ3D/ https://pretalx.adass.org/adass2026/talk/XRSQ3D/feedback/ Banquet Hall Contributed Talk 2026-11-03T16:30:00+08:00 16:30 00:15 Spectroscopic binaries and multiple-star systems are key samples for studying stellar formation, evolution, and dynamical interactions. However, identifying such systems in large spectroscopic surveys remains challenging because of their rarity, complex blended spectral features, and the high cost of visual inspection. In this talk, I present a Human-AI hybrid approach for searching double-lined and triple-lined spectroscopic multiple systems in the LAMOST Medium-Resolution Spectroscopic Survey. The method combines the cross-correlation function (CCF), machine-learning classification, and human verification to improve the efficiency and reliability of candidate selection. I also compare several classifiers, including SVM, Transformer, DNN, CNN, and LSTM, under different training-data configurations, and evaluate their generalization ability using both observational and independent test sets. Finally, I introduce how the candidate catalogs, CCF features, machine-learning results, and multi-source information from Gaia and existing binary/multiple-star catalogs are integrated into the China-VO multi-star system archive. The platform is designed not only as a data repository, but also as a tool for scientific discovery: it supports candidate cross-identification, visualization of spectra and classification results, identification of newly discovered systems, selection of special targets, and prioritization for follow-up observations. By combining large-scale survey mining with an online research platform, this work provides a practical pathway from spectroscopic candidate detection to scientific validation of multiple-star systems. adass2026-90-mining-double-line-spectroscopic-candidates-in-the-lamost-medium-resolution-spectroscopic-survey-using-a-human-ai-hybrid-method AI as tool for scientific discovery Shanshan Li en false https://pretalx.adass.org/adass2026/talk/FLP3BT/ https://pretalx.adass.org/adass2026/talk/FLP3BT/feedback/ Banquet Hall Contributed Talk 2026-11-03T16:45:00+08:00 16:45 00:15 Recent advances in software and data-processing strategies have transformed the capabilities of the LOw Frequency ARray (LOFAR), enabling wide-field radio imaging at unprecedented depth and angular resolution. In this talk, I will present the developments that have made ultra-deep sub-arcsecond imaging with LOFAR both computationally feasible and scientifically transformative. I will describe improvements to the LOFAR-VLBI imaging pipeline (PILOT), including automated self-calibration driven by signal-to-noise metrics, machine-learning-based image quality assessment, and optimised processing strategies that reduce computational costs by factors of 4-6 compared to previous approaches. To demonstrate these capabilities, we applied the updated pipeline to 200 hours of LOFAR High-Band Antenna observations of the ELAIS-N1 field, which corresponds to processing 400 TB of visibility data. The resulting image reaches a central sensitivity of 5.6 μJy beam⁻¹ at 0.3 arcsec resolution over a 2.5×2.5 deg² field, making it the deepest radio image below 3 GHz and one of the deepest radio images ever produced. This dataset opens a new window on the faint radio-source population, from distant active galactic nuclei to some of the earliest star-forming galaxies. More broadly, it demonstrates how advances in calibration, automation, and computational efficiency are paving the way for the next generation of ultra-deep radio surveys with LOFAR 2.0 and eventually the Square Kilometre Array (SKA). adass2026-75-pushing-lofar-to-its-limits-automated-ultra-deep-sub-arcsecond-radio-imaging-below-200-mhz Topical Computing, Software and Algorithms /media/adass2026/submissions/UWABRY/fullFoV_layout_200hrs_28U8bOc_Hrqj1ZS.webp Jurjen en false https://pretalx.adass.org/adass2026/talk/UWABRY/ https://pretalx.adass.org/adass2026/talk/UWABRY/feedback/ Poster Rooms Poster Session 2026-11-03T10:00:00+08:00 10:00 01:00 Poster session 3 adass2026-203-poster-session-3 Poster Andreas Wicenec en false https://pretalx.adass.org/adass2026/talk/SB9APV/ https://pretalx.adass.org/adass2026/talk/SB9APV/feedback/ Restaurant Welcome 2026-11-03T18:30:00+08:00 18:30 04:00 Conference Dinner The Point 306 Riverside Drive, East Perth, WA 6004 adass2026-204-conference-dinner General Andreas Wicenec en false https://pretalx.adass.org/adass2026/talk/UADAMT/ https://pretalx.adass.org/adass2026/talk/UADAMT/feedback/ Banquet Hall Invited Talk 2026-11-04T09:00:00+08:00 09:00 00:30 Placeholder adass2026-211-changhua-s-talk Building and operating science platforms and workflows in the Petabyte Era Changhua Li en false https://pretalx.adass.org/adass2026/talk/W9UZWE/ https://pretalx.adass.org/adass2026/talk/W9UZWE/feedback/ Banquet Hall Contributed Talk 2026-11-04T09:30:00+08:00 09:30 00:15 The Nancy Grace Roman Space Telescope will produce approximately 20 PB over its five-year mission, requiring data reduction systems that are elastic, reproducible, cost-aware, and tightly integrated with the Roman Research Nexus science platform. We present the architecture and operational model for the Space Telescope Science Institute’s Roman data processing pipeline. This fully cloud-based data processing system is part of the larger distributed ground systems developed for NASA's latest flagship astrophysics telescope. The Roman pipeline is built around containerized processing workloads orchestrated by Kubernetes and Apache Airflow, with fully reproducible infrastructure through Infrastructure as Code. The design supports multiple concurrent versions of the Roman calibration software, enabling controlled reprocessing, validation, and operational flexibility as algorithms and mission needs evolve. The system scales elastically from zero workers when idle to thousands of parallel workers during processing campaigns. We describe the architectural patterns used to support petabyte-scale processing, including workflow abstraction, storage considerations, external interfaces for data receipt and distribution, and access patterns for downstream systems. Observed load has exceeded thousands of concurrent jobs without issue. At demonstrated scale, the system easily provisioned 60TB of memory across 7,000 vCPUs achieving remarkable hourly throughput. These figures represent tested operating points rather than architectural limits; continued load testing is expected to explore far higher levels of concurrency, with practical scaling constraints driven primarily by cloud resource availability, service quotas, and cost. The presentation will also address design tradeoffs for speed, portability, scale, security, and cost, along with monitoring and observability strategies required to operate a large production cloud workflow system. We will also present lessons learned from operating at scale in AWS, including service limits, performance bottlenecks, and discuss how pipeline outputs integrate with the Roman Research Nexus to provide a compute-to-data science platform for accessible, reproducible analysis. adass2026-148-building-a-cloud-native-petabyte-scale-pipeline-for-the-roman-space-telescope Building and operating science platforms and workflows in the Petabyte Era John Glorioso en false https://pretalx.adass.org/adass2026/talk/ATU73E/ https://pretalx.adass.org/adass2026/talk/ATU73E/feedback/ Banquet Hall Contributed Talk 2026-11-04T09:45:00+08:00 09:45 00:15 Fornax Science Console is a NASA-funded, open-source platform that brings compute to astrophysics data and makes production-grade deployments repeatable for teams of any size. We present a "0 → hero" reference stack that uses Terraform (managed with Terramate) to provision an AWS-backed Kubernetes environment (EKS + FSX (OpenZFS)) and Flux for GitOps-based continuous reconciliation. This repeatable IaC enables teams to go from repository to production-ready cloud environment with the same code paths used for day‑2 operations. The platform couples interactive analysis tooling (JupyterHub) with a full operational observability suite — Grafana for dashboards, Loki and TimescaleDB for logs and metrics — and end-to-end cost visibility and alerting utilizing derived metrics. Platform configuration and operational policies are continuously enforced through GitOps, lowering time-to-first-analysis while reducing operational load by "eating our own dog food." We also describe an integrated support loop: a deployable Discourse instance that captures troubleshooting and how‑to content as searchable, citable knowledge base to scale user support. Attendees will leave understanding how our open platform can be for building reproducible compute‑near‑data platforms, running GitOps-driven operations, achieving end‑to‑end cost and reliability visibility, and cultivating a sustainable user support ecosystem that scales with the science. adass2026-130-fornax-science-console-a-repeatable-open-science-platform-for-compute-near-data Building and operating science platforms and workflows in the Petabyte Era Andrew Sawyers en I submitted a very similar abstract last year and was accepted and was unable to present due to US Government Shudown; the presentation/paper will be updated to reflect technical changes over the past 12 months. false https://pretalx.adass.org/adass2026/talk/LWJWHL/ https://pretalx.adass.org/adass2026/talk/LWJWHL/feedback/ Banquet Hall Contributed Talk 2026-11-04T11:00:00+08:00 11:00 00:15 The Square Kilometre Array Observatory (SKAO) is constructing the world's largest and most sensitive radio observatory, with SKA-Low in Australia and SKA-Mid in South Africa. The SKA telescopes will produce data volumes that require reliable processing pipelines capable of transforming raw visibility data into science-ready products such as continuum and spectral line image cubes and mechanisms to make the large data accessible to users across the world. While real-time processing supports telescope operations during observations and formation of coherent tied array beams, the computationally intensive calibration and imaging workflows are performed as “batch pipelines” after data acquisition. These workflows comprise multiple stages—including pre-processing, self-calibration, continuum imaging and spectral imaging—that must be orchestrated across heterogeneous compute and storage resources while maintaining reproducibility, robustness and operational efficiency. This paper focuses on the design challenges and emerging architecture of the SKAO Science Data Processor (SDP) batch pipeline framework, developed to support reliable, observatory-scale scientific workflows in the petabyte era.. The framework is intended to support a diverse range of observing modes through modular, configurable workflows that separate scientific processing from execution infrastructure. The SKAO is constructing the Low and Mid telescopes in stages called Array Assemblies (AA). The major array assembly stages are labelled as AA2, AA* and AA4. A key design objective is to provide an architecture that scales across successive processing capabilities as the observatory grows from AA2 through AA* to the full AA4. As computational and domain complexity increase across these capability levels, efficient management of visibility data becomes as important as the execution of the scientific algorithms themselves. These requirements have driven several key architectural decisions. An example is the adoption of the Measurement Set version 4 (MSv4) data model, developed in collaboration with the National Radio Astronomy Observatory (NRAO), that provides a data representation designed for parallel access to large visibility datasets. Together with workflow decomposition, configuration-driven execution, and well-defined interfaces between pipeline stages, these design choices establish a robust and extensible foundation for reliable batch processing while allowing calibration and imaging algorithms to evolve independently of the execution framework. Beyond addressing the immediate requirements of the SKAO, the proposed architecture establishes a foundation for reliable, modular and scalable batch processing that can evolve with the needs of future data-intensive astronomical observatories. adass2026-102-towards-reliable-batch-data-processing-for-the-ska-science-data-processor Building and operating science platforms and workflows in the Petabyte Era Ruta Kale en false https://pretalx.adass.org/adass2026/talk/RCLJJB/ https://pretalx.adass.org/adass2026/talk/RCLJJB/feedback/ Banquet Hall Contributed Talk 2026-11-04T11:15:00+08:00 11:15 00:15 ## Galactic Plane Imaging and Polarimetry from MeerKAT Data on High-Performance and Memory-Based Computing Resources **Authors:** L. Haupt¹, A. Basu¹, S. Borra¹, E. Buchholz¹, H. Heßling¹, Y. K. Ma² **Affiliation:** ¹ German Center for Astrophysics (DZA), Görlitz, Germany ² Max Planck Institute for Radio Astronomy (MPIfR), Bonn, Germany **Date:** July 28, 2026 --- ## Abstract The MPIfR–MeerKAT Galactic Plane Survey (MMGPS) produces large multidimensional imaging datasets, including full-polarisation image cubes and Faraday cubes. Processing these datasets requires high-capacity storage together with high-performance computing (HPC) resources, as image calibration and reconstruction demand exceptionally large amounts of memory. Unfortunately, processing the complete MMGPS workflow currently takes several weeks. At the German Center for Astrophysics (DZA), a multidisciplinary team with expertise in astrophysics, computer science, and mathematics performed a detailed analysis of the MMGPS processing pipeline with respect to memory, storage, and computational requirements. The study revealed significant I/O bottlenecks that limit the overall processing performance. Our objective is to optimise resource allocation by mapping each pipeline stage to the hardware architecture best suited for its requirements, thereby significantly improving efficiency without modifying the underlying scientific software. Memory-Based Computing (MBC) plays a central role in this approach. Unlike conventional HPC systems, which are primarily processor-centric, MBC is designed around a large shared-memory architecture. The DZA operates an MBC prototype equipped with 48 TB of shared main memory, providing an ideal platform for memory-intensive radio astronomy applications. By combining HPC and MBC resources, we successfully reconstructed a complete three-dimensional MeerKAT image cube. Furthermore, the large memory capacity of the MBC system enabled the generation of a Faraday cube and its direct visualisation as a movie. Overall, the optimised mapping of pipeline stages to the appropriate hardware improved processing performance by approximately one order of magnitude. In addition to these hardware-related improvements, we present first considerations for potential software-level optimisations. Future work will focus on identifying remaining bottlenecks and evaluating parallelisation strategies. The MMGPS pipeline serves as an ideal prototype for preparing data-processing workflows for the substantially larger datasets expected from the Square Kilometre Array Observatory (SKAO). adass2026-108-galactic-plane-imaging-and-polarimetry-from-meerkat-data-on-high-performance-and-memory-based-computing-resources Building and operating science platforms and workflows in the Petabyte Era Lars HauptElsa Buchholz (DZA) en false https://pretalx.adass.org/adass2026/talk/CNYTGX/ https://pretalx.adass.org/adass2026/talk/CNYTGX/feedback/ Banquet Hall Contributed Talk 2026-11-04T11:30:00+08:00 11:30 00:15 Data Fusion involves combining data from multiple facilities (x-ray through optical and infrared to radio), to understand the total energy output of astronomical objects and galaxies in particular. This data fusion process is vital to estimate robust distances (e.g., photometric redshifts), estimate accurate stellar masses (and other physical properties), and to assess the dynamical state of the objects we’re studying (e.g., merger and structure). However, to perform data fusion on millions of objects over hundreds of thousands of square degrees, requires addressing issues such as mismatched resolutions and sensitivities, footprint coverage and variable point-spread functions. Critically producing reliable robust measurements cannot be achieved in catalogue space, and but needs to be done at the pixel level. With many modern facilities now advocating a “bring your compute to the data” approach, this may actively present a serious barrier to multi-facility science. In this talk we will illustrate the problems and barriers, highlight how important data fusion is through examples from the Galaxy And Mass Assembly (GAMA) and Wide Area VISTA Extragalactic Survey (WAVES), and advocate for a global international capbility that can bring diverse pixels to the compute to enable data fusion from x-ray (eROSITA) to ultraviolet (GALEX), optical (LSST), infrared (Roman, Euclid, WISE, Herschel) and radio (ASKAP, MWA). adass2026-84-data-fusion-what-is-it-and-why-bringing-compute-to-the-data-may-endanger-it Building and operating science platforms and workflows in the Petabyte Era Simon Driver en false https://pretalx.adass.org/adass2026/talk/NQTYNR/ https://pretalx.adass.org/adass2026/talk/NQTYNR/feedback/ Banquet Hall Contributed Talk 2026-11-04T11:45:00+08:00 11:45 00:15 SKA Regional Centres operating at modest scale face a distinct set of challenges in the petabyte era: not how to store everything, but how to extract maximum scientific value from the data and hardware they do have. The Swiss SRC (CHSRC), hosted at CSCS, approaches this by leaning into its strengths rather than attempting to replicate larger node capabilities. With GraceHopper nodes available at CSCS, AI/ML-intensive science use cases are a natural focus, reflecting the priorities of the Swiss astronomy community — but integrating specialised HPC hardware into a science platform is itself a non-trivial architectural challenge. For data-hungry use cases such as EoR that will require compute-to-data across the SRCNet, the architecture must support workflows that can move between nodes, though this model is still ahead of us as we work towards SKAO Science Verification. This talk is grounded in CHSRC's preparation for SV: qualifying as an SV node, ensuring Swiss users can access SV data, and building a software platform that makes the most of the infrastructure available. It explores the architectural consequences of operating under fixed hardware constraints, where flexibility must live in the software and platform layer — including FirecREST integration as a mechanism for abstracting heterogeneous HPC resource access. We also discuss user support as a first-class activity in building the habits and tooling that enable reproducible, scalable astronomy. The Swiss case offers a perspective on how small partners can contribute meaningfully to a global infrastructure by being precise about where they add value. adass2026-100-making-the-most-of-what-you-have-platform-strategy-and-science-verification-readiness-at-the-swiss-src Building and operating science platforms and workflows in the Petabyte Era /media/adass2026/submissions/FV99CZ/SKAO_chSRC_logo_colour_rgb_99lGu21_WmZ4CzY.webp Rohini Joshi en The talk is grounded in CHSRC's practical experience building towards SKAO Science Verification, but the broader themes — constraint-driven architecture, AI/ML positioning, compute-to-data versus data-to-compute, and user culture change — are likely to resonate across SRC nodes and national platforms of varying scales. false https://pretalx.adass.org/adass2026/talk/FV99CZ/ https://pretalx.adass.org/adass2026/talk/FV99CZ/feedback/ Banquet Hall Contributed Talk 2026-11-04T13:30:00+08:00 13:30 00:15 The Murchison Widefield Array (MWA) can offer a unique insight into the population of low-frequency radio transients in the southern sky due to its wide field of view and unique high-time resolution voltage capture capabilities. The petabyte-sized MWA archive contains raw-voltage observations spanning more than 7 years that can be scoured for objects evolving over time. However, the lack of tools to address the associated computational challenge impairs its potential for systematic discoveries. In this talk I will present the Breakthrough Low-latency Imaging with Next-generation Kernels (BLINK), a high time resolution imaging pipeline. It is a new software designed and developed from the ground up to run image-based fast transient searches at petaFLOPS or even exaFLOPS scale on modern supercomputing infrastructure. It also features a new image-based dedispersion algorithm called Streaming high-Time Resolution Imaging DEdispersion (STRIDE). This imaging pipeline achieves a 3700x speedup compared to a one based on WSClean, while the new dedispersion algorithm, STRIDE, reduces memory requirements by more than 97% in a typical wide-field search scenario. The software has been deployed on Pawsey's Setonix supercomputer to search for Fast Radio Bursts (FRBs) in the SMART dataset, a process that analysed 4 PB of voltage data, generating 21 PB of intermediate images, over the course of 6 months. I will present early technical and science results of the search, demonstrating petaFLOPS and petabytes scale computing is possible today. adass2026-101-blink-and-you-ll-miss-it-a-high-time-resolution-imaging-pipeline-for-fast-radio-burst-searches-with-the-murchison-widefield-array Building and operating science platforms and workflows in the Petabyte Era Cristian Di Pietrantonio en false https://pretalx.adass.org/adass2026/talk/KQFGAJ/ https://pretalx.adass.org/adass2026/talk/KQFGAJ/feedback/ Banquet Hall Contributed Talk 2026-11-04T13:45:00+08:00 13:45 00:15 # Bridging Kubernetes and Slurm for Transparent HPC Job Offloading **Authors:** S. Borra<sup id="fnref:1"><a href="#fn:1" class="footnote-ref">1</a></sup>, U. Canbolat<sup id="fnref:1"><a href="#fn:1" class="footnote-ref">1</a></sup>, M. Drobek<sup id="fnref:1"><a href="#fn:1" class="footnote-ref">1</a></sup>, L. Haupt<sup id="fnref:1"><a href="#fn:1" class="footnote-ref">1</a></sup> **Affiliation:** <sup id="fnref:1"><a href="#fn:1" class="footnote-ref">1</a></sup> German Center for Astrophysics (DZA), Görlitz, Germany **Date:** July 30, 2026 **Abstract:** Kubernetes has emerged as a control plane for modern scientific platforms, supporting reproducible and high-throughput computing workflows. Additionally, high-performance computing (HPC) systems managed by Slurm remain essential for large-scale, compute-intensive processing. Despite their complementary roles, these environments are typically disconnected, requiring manual workload transitions and resulting in fragmented observability and control. Several approaches were explored to bridging this gap, including systems such as interLink, Volcano, REANA (Reusable Analyses) and Slinky. In this work, we present interLink, a lightweight integration that extends Kubernetes orchestration into a Slurm-managed HPC cluster. By representing HPC resources within Kubernetes, workloads can be offloaded directly and executed as Slurm jobs without requiring changes to user workflows or tooling. A Kubernetes pod, the smallest deployable unit in Kubernetes, encapsulating one or more containers and their execution context, acts as the interface through which users submit and monitor jobs. interLink ensures that job state, logs, and execution progress are continuously reflected back into this originating pod, preserving a consistent operational view. This approach enables Kubernetes to function as a unified control plane for both service-oriented and batch-oriented workloads. In particular, a remote user can use Jupyter notebooks interactively, for example, in preparation of batch jobs to be run on large datasets stored in the HPC center. This presentation explains how interLink reduces operational friction, maintains end-to-end observability across system boundaries, and enables seamless integration of high-throughput and HPC workflows within a single environment. adass2026-118-bridging-kubernetes-and-slurm-for-transparent-hpc-job-offloading Building and operating science platforms and workflows in the Petabyte Era Sagar Borra en false https://pretalx.adass.org/adass2026/talk/CTG9SR/ https://pretalx.adass.org/adass2026/talk/CTG9SR/feedback/ Banquet Hall Contributed Talk 2026-11-04T14:00:00+08:00 14:00 00:15 Astronomy is firmly entrenched in the big data era, with modern surveys routinely generating terabyte-scale image cubes. The visualisation and inspection of these massive datasets present significant challenges—bottlenecks that will only intensify with the advent of petabyte-scale facilities such as the SKA, ngVLA, and LSST. The Cube Analysis and Rendering Tool for Astronomy (CARTA) was developed to overcome these barriers. CARTA’s client/server architecture empowers scientists to remotely inspect massive datasets in situ, operating with a memory footprint far smaller than the original data size. In this talk, I will outline the specific challenges associated with visualising TB-scale image cubes and demonstrate how CARTA achieves smooth, interactive inspection along different dimensions. This is achieved, in part, by converting standard FITS files into a collection of datasets (currently saved in HDF5 format) tailored for efficient data retrieval. For instance, CARTA enables rapid spectral inspection by generating a "rotated" dataset optimised for reading along the frequency axis. I will discuss existing and planned solutions to further optimise the FITS-to-HDF5 conversion process, significantly reducing the time required to deliver inspection-ready data. Finally, I will highlight ongoing efforts to implement a multi-server architecture to visualise datasets stored in different locations across the globe. adass2026-114-visualising-massive-astronomical-datasets-with-carta Building and operating science platforms and workflows in the Petabyte Era Marcin Sokolowski en false https://pretalx.adass.org/adass2026/talk/WPUEXM/ https://pretalx.adass.org/adass2026/talk/WPUEXM/feedback/ Banquet Hall Contributed Talk 2026-11-04T14:15:00+08:00 14:15 00:15 We archive massive datasets to map the stars in ESA’s Gaia mission, unlock the secrets of dark matter and dark energy with Euclid, and discover new worlds with PLATO. Over the past few years, these workloads ran on Greenplum, but when Greenplum stopped being open source, we had to evaluate all available options—open source or not—exploring new directions. In this talk, we’ll share how we’re rethinking architecture, tools, and operations to handle petabyte-scale astronomy workloads. We’ll cover the practical pros and cons of the candidate solutions we evaluated, the administrative shifts required, and the hurdles we face moving such critical operations between engines. We have selected a solution that offers both Open Source and Enterprise support options. We'll discuss how this option emerged as the best fit for our petabyte-scale workloads, including its ability to handle complex cross-match queries essential for astronomical research. Attendees will leave with real-world lessons from our migration journey, including performance benchmarks from Gaia catalog queries, operational considerations for distributed PostgreSQL at scale, and key questions to ask before starting a migration of their own. adass2026-142-from-stars-to-storage-engines-migrating-big-science-workloads-beyond-greenplum Building and operating science platforms and workflows in the Petabyte Era Joaquim Oliveira en false https://pretalx.adass.org/adass2026/talk/GNKDSF/ https://pretalx.adass.org/adass2026/talk/GNKDSF/feedback/ Banquet Hall Invited Talk 2026-11-04T14:30:00+08:00 14:30 00:30 Placeholder adass2026-212-elizabeth-s-talk Usability, accessibility and security in astronomy software Elizabeth Davies en false https://pretalx.adass.org/adass2026/talk/NNDZSY/ https://pretalx.adass.org/adass2026/talk/NNDZSY/feedback/ Banquet Hall Contributed Talk 2026-11-04T16:00:00+08:00 16:00 00:15 Documentation is one of the primary ways that users interact with scientific software, yet it is often treated as an afterthought rather than an integral part of the software itself. Well-designed interactive documentation can reduce barriers to entry, improve reproducibility, and enable researchers to quickly adopt new tools and workflows. This talk presents practical strategies for developing interactive documentation for astronomy software, drawing on my experience maintaining documentation for multiple Python packages and MAST's repository of Jupyter notebooks. I will discuss how executable notebooks, narrative tutorials, and example-driven documentation can guide users from introductory concepts to real scientific workflows. I will also share approaches for testing documentation alongside code to ensure examples remain functional and trustworthy. In addition, I will explore how AI-assisted development can support documentation authoring and maintenance. Rather than replacing human expertise, AI tools can help draft tutorials, improve clarity, identify knowledge gaps, and update examples as APIs change, allowing developers to spend more time refining content and less time on repetitive maintenance. By treating documentation as a core component of scientific software rather than an afterthought, astronomical software projects can improve usability, lower the learning curve for new users, and increase the long-term impact of their tools. adass2026-144-interactive-documentation-in-the-age-of-ai-lessons-from-mast-software Usability, accessibility and security in astronomy software Sam Bianco en This talk would be relevant to multiple conference themes. false https://pretalx.adass.org/adass2026/talk/GPAYMV/ https://pretalx.adass.org/adass2026/talk/GPAYMV/feedback/ Banquet Hall Contributed Talk 2026-11-04T16:15:00+08:00 16:15 00:15 The ProposAl Workflow System (PAWS), a web-based tool, is used to manage proposal submissions for the MeerKAT Radio Telescope, making reliability an important part of the user experience.This work describes the introduction of automated end-to-end testing, using the Playwright framework, to verify proposal submission workflows from the perspective of the end user. The test suite was used to automate the creation of draft proposals in a test batch and exercises key stages of the submission process, including team information, science case details, target selection, observation configuration, data management, PDF uploads and the final submission stage. Stable ‘data-testid’ attributes were added to improve test reliability, while reusable PDF fixtures enabled consistent validation of document uploads. The tests also covered common validation scenarios including required fields, invalid email addresses, upload behaviour and warnings for publicly accessible Google Drive links. In addition to increasing confidence in the proposal workflows, this work establishes a foundation for broader automated testing including proposal constraints, permission based behaviour and more comprehensive submission validation. As a junior developer, implementing these tests also proved to be an effective way to understand the PAWS codebase while making a practical contribution to its quality and maintainability. adass2026-86-building-confidence-in-the-proposal-submission-system-with-playwright Usability, accessibility and security in astronomy software Ameera Gangat en false https://pretalx.adass.org/adass2026/talk/RZCDHF/ https://pretalx.adass.org/adass2026/talk/RZCDHF/feedback/ Banquet Hall Contributed Talk 2026-11-04T16:30:00+08:00 16:30 00:15 Modern astronomical web platforms increasingly integrate interactive data exploration, authenticated data access, and online analysis services. As these capabilities expand, usability, accessibility, and security become fundamental design requirements. We present the Cherenkov Telescope Array Observatory (CTAO) Science Data Challenge (SDC) Portal as a case study of how these complementary requirements can be addressed within a modern, open-source astronomical web platform. The portal couples a React front end with a FastAPI microservice architecture and exposes observation metadata through International Virtual Observatory Alliance (IVOA) standards, including TAP/ADQL and ObsCore. These community standards make Cherenkov data accessible to astronomers beyond the high-energy community while promoting FAIR access to astronomical data through open interfaces and an open-source implementation. Usability is supported through source-name resolution, coordinate and time conversions, interactive visualisations, reusable data baskets, and submission of selected observations as preview jobs. Accessibility is improved through colour-blind-safe visualisations and CTAO usability guidelines. The platform follows a service-oriented architecture separating authentication, application logic, and data download. Authentication and authorization are handled through OpenID Connect. A dedicated download service resolves logical file names to distributed storage endpoints and performs OAuth 2.0 Token Exchange for direct file downloads. Persistent application data are associated only with pseudonymous user identifiers, while personal profile information remains confined to temporary session state. The platform is deployed using containers, Helm, and Kubernetes, supported by continuous integration, automated code quality checks, and Grafana observability. Together, these engineering practices improve software quality and maintainability, while ongoing cybersecurity hardening prepares the platform for future CTAO operations. adass2026-96-the-ctao-science-data-challenge-portal-a-secure-and-user-centric-web-platform-for-gamma-ray-astronomy Usability, accessibility and security in astronomy software oates en false https://pretalx.adass.org/adass2026/talk/WJLXPM/ https://pretalx.adass.org/adass2026/talk/WJLXPM/feedback/ Banquet Hall Contributed Talk 2026-11-04T16:45:00+08:00 16:45 00:15 Most astronomical machine learning workflows begin only after astronomy specific processing has already taken place: selecting archive products, interpreting FITS metadata, transforming coordinates, matching catalogues, resampling exposures, and combining observations across filters. These operations usually sit outside the model, making them difficult to optimize jointly, test within the same computation, or execute efficiently on GPUs. We present torchfits and torchsky, a tensor native software stack that connects FITS data with differentiable models of the observed sky. torchfits provides selective access to local and remote FITS images and tables, exposing tensors and columnar batches for analysis and training. torchsky provides celestial geometry, catalogue association, mapmaking, spectral response, PSF convolution, and other observation operators on a common PyTorch runtime, with execution on CPUs and GPUs. Together, these tools support multispectral representations of the sky as a function of position and wavelength, fitted to heterogeneous observations rather than restricted to precomputed coadds or catalogues. The same components also support conventional survey processing and large-scale cross-matching. We will describe the software architecture, numerical validation, performance, and initial applications to multispectral mapmaking and survey-scale catalogue association. We will also show how torchsky’s geometric primitives are being reused in a new cross-matching system. More broadly, placing data access, celestial geometry, and observation operators on a common tensor runtime opens new possibilities for joint inference across surveys and wavelengths, differentiable calibration and mapmaking, and foundation models that learn from astronomical observations rather than only from preprocessed products. adass2026-162-from-fits-archives-to-differentiable-multispectral-sky-models Topical Computing, Software and Algorithms Sébastien Fabbro en false https://pretalx.adass.org/adass2026/talk/UYVFZS/ https://pretalx.adass.org/adass2026/talk/UYVFZS/feedback/ Banquet Hall Contributed Talk 2026-11-05T09:30:00+08:00 09:30 00:15 A major obstacle in time-domain and multi-messenger astronomy is the high entry barrier caused by a fragmented follow-up ecosystem. While separate tools exist for alert brokering, telescope triggering, and data reduction, investigators must navigate disparate software interfaces, limiting the speed of rapid follow-up campaigns. Addressing this coordination gap requires a unified, end-to-end framework. To solve this, Gemini/NOIRLab has built the Gemini Observation and Analysis of Targets System (GOATS). This single web application consolidates the entire workflow into a responsive chain: as new targets are ingested, Target-of-Opportunity (ToO) observations are automatically triggered via the new Gemini Program Platform (GPP), data are automatically retrieved from archives, and data reduction is performed directly in-browser with live status updates. GOATS further includes lightweight image analysis tools and features to publish final data products to external services, enabling researchers to transition swiftly from initial alert to calibrated data. This workflow was exercised in a live end-to-end demonstration driven by real Rubin alerts, with ToOs submitted to Gemini North and South, and other NOIRLab facilities. This presentation will cover the primary components, architecture, and practical challenges of engineering this multi-service workflow. I will discuss key elements, which include a browser extension designed to convert alerts into targets, the adaptation of a CLI/Python data reduction pipeline into an interactive web UI, the infrastructure for streaming progress from long-running tasks, and programmatic integration with the GPP. I will also outline ongoing efforts to bring interactive, in-browser data visualization and analysis into the same unified workflow. Ultimately, GOATS simplifies a complex landscape into a dependable platform, lowering the barrier to rapid follow-up so scientists can focus on research rather than tool integration. adass2026-128-goats-an-end-to-end-time-domain-and-multi-messenger-astronomy-platform Topical Computing, Software and Algorithms Miguel Gómez en false https://pretalx.adass.org/adass2026/talk/S9FKAW/ https://pretalx.adass.org/adass2026/talk/S9FKAW/feedback/ Banquet Hall Contributed Talk 2026-11-05T09:45:00+08:00 09:45 00:15 The friends-of-friends (FoF) algorithm has been used for over 40 years in extragalactic astronomy to identify galaxy-groups in redshift surveys. Whilst this simple percolation algorithm was sufficient in the era of 100s of thousands of galaxies, the next generation of redshift surveys (DESI, WAVES, 4HS) promise to take us well into the 10s of millions. And, since FoF scales quadratically, we have reached a point where the classic implementation is entirely impractical. **Nessie** is a group finder with the explicit goal of modernizing the FoF algorithm for this next generation of redshift surveys. In this talk I will explain how Nessie implements tree data structures to reduce the complexity of the FoF algorithm; why this is different to other group finders and other FoF implementations; and finally, talk about how the rust ecosystem made Nessie possible, and attempt to convince other astronomers to adopt it for their next project. adass2026-76-nessie-the-fastest-group-finder-in-the-world-or-why-rust-is-the-best-high-level-language Topical Computing, Software and Algorithms /media/adass2026/submissions/33MGVM/1_FJ0dNPY_9PhwpB0.webp Trystan Lambert en false https://pretalx.adass.org/adass2026/talk/33MGVM/ https://pretalx.adass.org/adass2026/talk/33MGVM/feedback/ Banquet Hall Contributed Talk 2026-11-05T11:15:00+08:00 11:15 00:15 Spectroscopic data require accurate wavelength solutions to fully unlock their scientific potential. While there are many tools for fitting wavelength solutions, there are no tools tailored specifically for solar data. In this talk we discuss the development of a new Python package called `solar-wavelength-calibration`, maintained by the Daniel K. Inouye Solar Telescope's Data Center and designed to accurately fit wavelength solutions to spectroscopic observations of the Sun. The novel features of this package include the use of a solar spectral atlas as opposed to line lists (no need for arc-lamp data!) and the explicit treatment of telluric lines with varying intensity and an optional Doppler offset. Additionally, the parametrization of the wavelength solution does not use polynomials, as is common in other tools, but rather relies on a physically motivated model, with fit parameters corresponding to the optical characteristics of the spectrograph. As a result, the final fit is fully compliant with the FITS parametrization of non-linear wavelength axes for grating data (AWAV-GRA), eliminating the need to interpolate data onto a linear grid. `solar-wavelength-calibration` is available on pypi and published in JOSS. This talk will give a brief overview of how this package is used in the wild, but will focus mostly on lessons learned and the process of turning a collection of functions that work on specific data into a general purpose tool for astronomers everywhere. adass2026-129-solar-wavelength-calibration-designing-a-new-tool-for-computing-wavelength-solutions-of-solar-spectra Topical Computing, Software and Algorithms Arthur Eigenbrot en false https://pretalx.adass.org/adass2026/talk/NWTBRJ/ https://pretalx.adass.org/adass2026/talk/NWTBRJ/feedback/ Banquet Hall Contributed Talk 2026-11-05T11:30:00+08:00 11:30 00:15 Fiber-fed spectrographs now dominate ground-based astronomy. Integral-field units map galaxies point by point; multi-object systems place thousands of fibers across a field; the facilities of the next decade will multiply both. Every one of these instruments needs the same reduction: trace the fibers, calibrate wavelength and flux, remove the sky, and assemble the results into spectra or cubes. The software foundation has not kept pace. Much of the community still depends on IRAF, which is no longer developed, is written in a language few can maintain, and requires interactive steps that cannot be scripted or reproduced. The alternative has been a separate pipeline for each instrument, duplicating effort and retiring with the hardware. Meanwhile, expectations have risen: surveys must reduce data unattended for years, archives must reproduce results long after publication, and students must be trained on tools with a future. The gap between what fiber spectroscopy needs and what its software provides is growing. We developed FIBRA, a Python pipeline that reduces fiber-fed spectroscopy from raw detector frames to calibrated data cubes. One code base serves different instruments: an instrument is described by data files, not by changes to the pipeline. Calibration steps that traditionally required an expert at a cursor run automatically and report their own confidence. Sky subtraction follows instrument signatures across the field instead of assuming a uniform sky. Uncertainties are carried through every stage, and cubes are built within the memory of an ordinary computer. Each step can be run in batch or tuned interactively, and both produce the identical result from the same configuration. Every product records how it was made. The documentation is verified by the software itself and an illustrated guide takes a newcomer from raw data to a first measurement. The talk will present the design and walk through a complete night of data reduced with the pipeline, from raw frames to a science measurement. We will focus on the decisions that matter beyond this one code: how a single pipeline serves many instruments, how interactive work is made exactly repeatable, and what this experience suggests for the teams now building reduction systems for the coming generation of fiber-fed facilities. adass2026-83-fibra-fiber-bundle-reduction-of-astronomical-spectra-an-instrument-agnostic-pipeline-for-fiber-fed-integral-field-spectroscopy Topical Computing, Software and Algorithms Sabyasachi Chattopadhyay en true https://pretalx.adass.org/adass2026/talk/DXBJFU/ https://pretalx.adass.org/adass2026/talk/DXBJFU/feedback/ Banquet Hall Contributed Talk 2026-11-05T11:45:00+08:00 11:45 00:15 As part of the EU-funded RADIOBLOCKS project, we have created a set of highly optimized GPU libraries which can serve as building blocks for a correlator or beamformer application. We have integrated a number of these radioblocks into a fully functional VLBI correlator, which is intended to become the next production correlator of the European VLBI Network (EVN). In this talk I will present this correlator and its constituent radioblocks. Since 2012, the operational correlator of the EVN is a distributed CPU based correlator which is deployed locally on an HPC cluster. However, the advent of next generation wideband receiver systems is expected to increase the maximum observed bandwidth in EVN observations by a factor four to eight, which will require an increase in correlator capacity by a similar amount. Expanding the current HPC cluster to meet these new requirements would be prohibitively expensive while also using a significant amount of power. However, as I will show in this talk, these requirements can be met cost-effectively using GPU acceleration, while drawing significantly less power. In the talk I will present detailed measurements of the power consumption of our GPU implementation. The GPU correlator is not a monolithic application; it makes use of a collection of GPU libraries that we have created, which we call radioblocks. Each of these radioblocks is distributed as a stand-alone library, allowing other projects to easily incorporate the relevant radioblocks into their own code base. For example, a core component of a VLBI correlator is the geometric delay compensation for which we have created an optimized VLBI capable radioblock. The (cross-)correlations are performed using a radioblock called the tensor core library (TCC), which uses tensor core instructions to compute the correlation function. It achieves an order-of-magnitude performance improvement over non-tensor-core GPU implementations. Furthermore, we have implemented a number of advanced features as radioblocks. We have created a radioblock which can perform coherent de-dispersion on baseband data, this feature is crucial to process certain classes of radio transients such as Fast Radio Bursts (FRBs), and some pulsars. Another example is a radioblock which can produce multiple simultaneous phase centres, a technique to efficiently perform wide-field VLBI. It does this by producing individual narrow-field data sets for each source in the field of view, rather than producing a single monolithic wide-field dataset that encompasses the entire field of view. I will conclude the talk by discussing a potential future application. The Space Array initiative, of which we are members, is preparing a proposal to build a space based VLBI array. Due to bandwidth limitations, the correlation will be performed onboard of the spacecrafts. I will show how our work could be applied to this space-based correlator. adass2026-131-sfxgpu-a-gpu-accelerated-correlator-for-the-european-vlbi-network Topical Computing, Software and Algorithms Aard Keimpema en false https://pretalx.adass.org/adass2026/talk/QL8JSB/ https://pretalx.adass.org/adass2026/talk/QL8JSB/feedback/ Banquet Hall Contributed Talk 2026-11-05T12:00:00+08:00 12:00 00:15 In this contribution we present a collection of highly optimized GPU signal-processing libraries that form the building blocks for a new generation of radio-astronomical correlator and beam-former applications. This collection was developed by several partners in the EU-funded RADIOBLOCKS project. It includes, amongst others a PolyPhase filter bank, the Tensor-Core Correlator, a Tensor-Core Beam Former and a VLBI-capable delay correction module. We describe how recent AI driven GPU hardware developments, as well as developments in the software ecosystem, have made it possible to achieve significant improvements in power-efficiency for these building blocks. We discuss briefly how these "radio blocks" will be deployed in (upgrades of) correlator/beam-former applications for a variety of radio telescopes such as LOFAR, the Effelsberg radio telescope, the Irbene Single Baseline Interferometer (ISBI) and the European VLBI Network (EVN). Finally we provide some practical examples of the programming techniques that we've used that could be used in other applications. adass2026-80-radioblocks-using-gpus-for-signal-processing-in-the-ai-era Topical Computing, Software and Algorithms Mark Kettenis en false https://pretalx.adass.org/adass2026/talk/UTWRDM/ https://pretalx.adass.org/adass2026/talk/UTWRDM/feedback/