RSE.swiss - The Swiss RSE community – Annual Swiss RSE day 2026

Accelerate cosmological simulations

Mon, 01 Jan 0001 00:00:00 +0000

Presenter: Tommaso Boschi from ETH Zurich

This talk is designed for RSE with an interest in high-performance computing (HPC) and GPU acceleration. No particular skills are required to understand the presentation, but familiarity with C++/CUDA and GPU-parallel algorithms will help fully engage with the technical details.

I will address the computational bottleneck in radiative transfer cosmological simulations, specifically the ray-tracing algorithm for calculating hydrogen column density along ionising photon paths. To overcome this, we have updated pyC2Ray, a modern, massively parallel successor to the widely used C2Ray library. pyC2Ray implements the Accelerated Short-characteristics Octahedral RAy-tracing (ASORA) algorithm, a novel GPU-optimised method for multi-source ray-tracing. The library combines C++/CUDA for performance with a Python interface for usability, enabling customisation without sacrificing efficiency. Compared to C2Ray, the updated version of pyC2Ray now achieves a 800x speed-up, demonstrating the transformative potential of GPU acceleration and specialised software design in cosmological simulations. This work highlights how targeted algorithmic and architectural optimisations can unlock new scales of computational efficiency and physical accuracy.

An open-source chat platform for independent research: Matrix at EPFL

Mon, 01 Jan 0001 00:00:00 +0000

Presenter: Carine Dengler from EPFL

Independence is one of the cornerstones of the research ecosystem, and as research software engineers we can support independent research through the tools we pick. Another core pillar is communication, which is why the research software engineers from the Center for Digital Trust (C4DT) and the Center for Imaging set up matrix.epfl.ch for EPFL’s research community.

Matrix is an open-source, open-protocol, secure and decentralised chat platform, which is the kind of solution universities need when they want to foster independent research. The service has been readily adopted by researchers and research software engineers alike, and together with the Matrix instance of our sister institution ETH Zurich played an important role in connecting the RSE communities across Switzerland.

In our talk we will walk you through the tech stack, show how we keep the service running - and evolving - on a tight budget, and talk about the community-building efforts to turn what started as a grassroots service into a sustainable, long-term solution.

Building Sustainable Scientific Software: Lessons from Industry–Academia

Mon, 01 Jan 0001 00:00:00 +0000

Presenter: Arnaud Mercier from PhD in Mathematical Geophysics in D-EAPS at ETH Zurich

Scientific software often faces a difficult balance between rapid research development and long-term maintainability. As new methods are introduced, ensuring reproducibility and reliability becomes increasingly challenging.

In a collaboration between the Mathematical Geophysics Group at ETH Zurich and Mondaic AG, we developed an open-source Python framework for full-waveform inversion that combines research flexibility with modern software engineering practices. Beyond the software architecture itself, a key focus of the project has been the development of testing strategies adapted to scientific applications.

While unit tests are widely adopted, they often fail to capture the complexity of real scientific workflows. We therefore complement them with integration tests that validate complete inversion pipelines on representative benchmark problems. These tests help ensure that new features, algorithmic changes, and performance improvements do not inadvertently alter scientific results.

This contribution discusses the challenges of maintaining a research codebase used for both methodological development and production-scale computations, highlighting lessons learned in modularity, testing, reproducibility, and collaboration between academia and industry. I would like to keep the discussion accessible and focused on general software engineering principles, development practices, and the lessons I learned through collaboration with a start-up.

Audience:
Research software engineers, scientific programmers, and researchers developing computational tools.

What participants will learn:<br/ Practical approaches to improving the reliability and sustainability of scientific software, with a particular focus on testing strategies that go beyond traditional unit tests.

Code Is Cheap Now — Understanding Users Isn't

Mon, 01 Jan 0001 00:00:00 +0000

Presenter: Ali Khosravi from PSI

Main points of your contribution:

The hard part of RSE isn’t the code, it’s understanding users and how they already work.
A real case study: a data-sharing platform for a 30+ group, and projected ~200-researcher Swiss consortium.
A concrete ‘‘understand → build → ask → repeat’’ loop, including real questionnaire findings.
A reflection on how AI shifts the value of the job: when prototypes take days, user understanding, communication, and ownership become the scarce skills.

What listeners will learn:

A practical method for designing tools researchers will actually adopt (not just tools that work).
Why existing habits make or break adoption, and how to account for them.
How to run the build–feedback loop in practice, with lessons from a live project.
How to rethink their own role, as AI makes code cheap.

Intended audience & skills needed:

Basically anyone who’s interested in building tools for researchers, there will also be some values for research group leaders to see the experience. No specialist background needed: no specific programming language, domain, or FAIR-data expertise required.

Connecting life cycle assessment expertise to research & development, proposal of an interactive digital evaluation and reporting tool

Mon, 01 Jan 0001 00:00:00 +0000

Presenter: Francesco Barilli from EMPA (SG)

INTENDED AUDIENCE
This presentation is intended for any kind of researcher interested in Life Cycle Assessment (LCA), knowledge formalisation, and interactive software tools. No prior knowledge of LCA is required. It may also be of interest to researchers exploring how AI can support the rapid development of prototype web-based applications.

MAIN CONTRIBUTION<br/ Life Cycle Assessment is a methodology used to evaluate the potential environmental impacts of products, technologies, and services. A complete LCA consists of four phases: Goal and Scope definition, Life Cycle Inventory Modelling and Computation, Impact Assessment, and Interpretation. The current focus of EMPA Life Cycle Assessment Support Centre (ELSC) is to help researchers with the implementation of the first phase, where they must define the scope, assumptions, and methodological choices that will influence the entire study and the next phases, a task that typically requires support from LCA experts.

ELSC thus created a web-based support tool that translates this expertise into an adaptive questionnaire, where previous answers influence subsequent questions . Through this interactive questionnaire, researchers are guided in a simple manner through the key information needed to define the study, including the purpose of the study, system boundaries, assumptions, and other methodological choices. So, the project explores how methodological guidance and expert knowledge in LCA can be translated into an adaptive questionnaire, bridging the gap between researchers with expertise in their own domains and the LCA expertise required to conduct transparent, reliable, and standards-compliant studies.

A second contribution of ELSC concerns the development of the interactive questionnaire itself. Indeed, we can show how AI-assisted development enabled a researcher with limited web-development experience to rapidly transform an Excel-based prototype of the LCA questionnaire into a maintainable and interactive web application that can be tested and iteratively improved with researchers.

WHAT LISTENERS WILL LEARN<br/ Listeners will see a live demonstration of the dynamic questionnaire and how it works through a case study. The demonstration will also explain how ELSC aims to support researchers during the Goal and Scope definition of an LCA by making methodological guidance more accessible through the questionnaire. They will also gain insight into how AI-assisted development can help domain scientists move from spreadsheets and static documents to web-based research software, accelerating prototyping, testing, and the rapid development of ideas that might otherwise require substantially more time and resources.

DALIA - High Performance Statistical Modelling made accessible

Mon, 01 Jan 0001 00:00:00 +0000

Presenter: Stefan zur Bonsen, Student at ETH Zurich

Dalia is a Python library implementing some statistical modelling tools that I have been working on as a semester project.

The library is currently being refactored to be more modular and less messy, so that it can be used by more people.

My part in this project is the implementation of the backend and making sure that the framework works on both CPUs and Accelerators (GPUs/NPUs). The poster would contain information about how that is done.

It should be understandable for anyone with a basic understanding of Python and numpy/scipy.

Readers will learn about hardware acceleration with cupy and nvmath as well as how to possibly unify a backend to support hardware acceleration and normal operation.

Danu - exploring federated services in your browser

Mon, 01 Jan 0001 00:00:00 +0000

Presenter: Linus Gasser from EPFL

Audience: web and backend developers with a knack for trying out obscure things - knowledge of WebRTC, Rust or TypeScript helps, but is not necessary

Main points: Yes we can break out of the walled gardens in the internet and collaborate using decentralised tools. However, the way of working in federated networks is different from the centralised approach learnt in most courses.

Learning: how to share trust and collaborate - how to set up a server in the danu network (just scan the QRCode :)

Defining Data Steward and RSE Roles Through a Task-Based Survey

Mon, 01 Jan 0001 00:00:00 +0000

Presenter: Stefanie Hauser from EMPA

Research Software Engineers (RSEs) and Data Stewards (DSs) have gained visibility in recent years, supported by emerging professional networks and increasing recognition in research institutions. At the same time, the boundaries between these roles remain unclear. Within the DSN-ORD and RSE4ORD projects, we aim to further define the roles of Data Steward and Research Software Engineer while strengthening the visibility of both profiles and supporting the development of their respective communities.

To contribute to this effort, we designed a task-based survey that compares the activities associated with both roles. In autumn 2025, we compiled an initial task list by reviewing online job advertisements for DS and RSE positions. The draft list was then reviewed by members of the DSN-ORD and RSE4ORD projects.

The questionnaire groups task statements into main categories and invites additional comments where needed. It takes less than 10 minutes to complete and will remain open until the end of July 2026. By identifying areas of overlap and difference, the survey is intended to support clearer job descriptions, more targeted career development, and a more evidence-based discussion of how DS and RSE roles relate to one another in research organisations.

Check out the questionnaire here: https://rse.swiss/blog/2026_06_09_ds_rse_task_questionnaire/

Developing career pathways for RSEs and dRTPs: A UK perspective

Mon, 01 Jan 0001 00:00:00 +0000

Presenter: Jeremy Cohen from Imperial College London

The term Research Software Engineering (RSE) was first used in the UK over a decade ago and it is now a recognised role within the field of research in many countries. As research across almost all domains is now increasingly digital, researchers need support not only with building sustainable, maintainable and robust research software, but also with managing and processing research data and the use of specialist research computing infrastructure. Within the UK, this has led to the emergence of a new term, ‘digital Research Technical Professionals’ (dRTPs) which is being used to represent the wider set of roles that are vital to supporting modern digital research.

While the UK now has RSE groups or teams at a number of universities and the RSE role is widely recognised, RSE roles differ and can span a wide range of skills and competencies. At one end of the scale, RSE roles can look quite similar to traditional research roles, at the other, they can look similar to professional software engineer roles in an industry environment. How we provide effective support and career progression for people working across this spectrum, and in other dRTP roles, is not yet a solved problem.

In this talk, I’ll present some work being done in the UKRI-EPSRC-funded STEP-UP project to develop support for RSEs and dRTPs within the London and South East of England region of the UK, through a range of activities. I’ll highlight our work advocating for improved career opportunities and pathways and our programmes to try and build concrete pathways into dRTP roles including a placement scheme and our Research Technical Champions programme. There is much that RSE communities across different countries can learn from each other. It is hoped that this talk and the associated discussion will provide an opportunity to share experiences and ideas around developing RSE career pathways and to understand where the Swiss and UK communities share similar challenges and what approaches and activities have been tried to address them.

This talk is intended to be applicable to a general RSE and research audience. Attendees will learn about some approaches being taken in the UK to support and develop career opportunities for RSEs but also have the chance to provide their own perspectives on the topic of RSE careers through an interactive element of the presentation and the subsequent discussion.

Developing software engineering 'industry skills' for graduates

Mon, 01 Jan 0001 00:00:00 +0000

Presenter: Ben Thomas from Software Sustainability Institute UK

The Software Sustainability Institute (SSI) had heard criticism from industry that research graduates do not join the workforce with the right software engineering skills. We wanted to explore whether research students and staff, including RSEs, benefitted from the skills that people had in industry.

Drawing on qualitative data from 26 interviews with employers, recent graduates, and industry stakeholders, we analysed perceptions of the skills graduates bring as well as skill gaps, and the causes of these gaps. We found that graduates were well-equipped with technical skills, but due to the way universities teach, often lacked the ability to apply skills practically, as well as lacking the professional (soft) skills needed for the workplace. While large employers were able to address skill gaps, small and medium-sized enterprises (SMEs) had more constrained resources, meaning they struggled to provide training, making workforce readiness more challenging in this vital part of the sector.

In response to these findings, the SSI training team developed a programme for undergraduate and research students which, through realistic projects, agile teamwork, mentoring, and reflection, helps early-career software engineers bridge the gap between academic study and professional practice. The programme has been piloted 3 times with University of Southampton students, receiving excellent feedback. Our next step is to offer this training to SMEs in the local area.

This research and training programme has placed the SSI in a strong position to shape future higher education policy around employability, and has already supported Southampton’s RSEs with implementing industry working practices.

Empowering sustainable research projects with Renku

Mon, 01 Jan 0001 00:00:00 +0000

Presenter: Elisabet Capon Garcia from Swiss Data Science Center

In this talk, we present Renku which stands for a trustworthy platform that enables collaboration and ensures reusability and repeatability of computational-based research projects. The objective is to raise awareness of the importance of project sustainability and lowering the entry barrier for reproducibility of AI and DS project by ensuring a smooth integration of data, code and compute infrastructure.

Listeners will get familiarised with the Renku platform, where they can create and share projects, and seamlessly connect them to their data sources and code repositories through a unified and user-friendly interface. Renku users do not need to handle the complexity to set-up the connections with their cloud providers because Renku handles this behind the scenes. Even more, the listeners will learn to set up different environments for their projects, from Python or R based development environments to Shiny or streamlit apps to showcase their results.

On the compute side, Renku offers a free tier with available compute resources to run sessions. If users require more compute resources than the free tier, they can either connect to their preferred cloud provider through Renku, connect to HPC clusters or contact us to provision further resources for them and their team.

EVERSE Project: Tools and Resources for Research Software Quality

Mon, 01 Jan 0001 00:00:00 +0000

Presenter: Kenneth Rioja from CERN

Abstract

EVERSE (https://everse.software/) is a community-led, EU-funded project bringing together 18 European institutions to establish a framework for research software quality and code excellence. Collaboratively designed by research communities, EVERSE aims to drive a cultural shift where research software is recognised as a first-class citizen of the scientific process and its contributors are duly credited. We will present the tools and resources EVERSE has developed over the last two years – ranging from practical guides and CLI tools to full platforms – tailored for researchers, RSEs, and those involved in research infrastructure or policy. We warmly welcome constructive feedback and early adopters.

Intended audience

RSEs, researchers, all levels are welcomed

Main contributions

A list of tools and resources to be shown and ready-to-be-used to the audience.

Learning objectives

What is a Research Software
What is EVERSE and its contributions
How to enhance my research software code quality through a guide I follow, or CLI/GitHub Actions
How I can get credited for the software work I did

Fractal: Towards FAIR bioimage analysis at scale with OME-Zarr-native workflows

Mon, 01 Jan 0001 00:00:00 +0000

Presenter: Lorenzo Cerrone from BioVisionCenter, Universität Zürich

The poster introduces Fractal, a platform for running FAIR bioimage analysis at scale using the OME-Zarr file format. It targets a few different audiences: people building large-scale image processing workflows in HPC contexts, people interested in Zarr and OME-Zarr, and developers of scientific platforms that bridge domain users and computational infrastructure.

Extended Abstract: Analysing large amounts of microscopy images in a FAIR manner is an ongoing challenge, made harder by the large diversity of image file formats and processing approaches. Recent work on OME-Zarr, a community-driven next-generation file format, offers the chance to create more shareable bioimage analysis workflows. At the BioVisionCenter, we are developing open-source resources for OME-Zarr-based image analysis. First, we propose extensions to OME-Zarr to incorporate tabular data for region of interest definitions, feature measurements and image metadata. Second, we have built a specification for interoperable image processing tasks that handle image data in the OME-Zarr format. Third, we have developed the Fractal framework to handle scalable and accessible image analysis workflows using OME-Zarrs. The Fractal framework consists of a server backend & web-frontend that handle modular image processing workflows. It facilitates the design and execution of reproducible workflows to convert images into OME-Zarrs and apply advanced processing operations to them at scale, without the need for expertise in programming or large image file handling. The Fractal community has made over 100 tasks publicly available under permissive licenses that enable converting vendor data to OME-Zarr, performing segmentation using a variety of segmentation networks, extracting high-dimensional measurements from large datasets, processing multiplexed images and many more analysis needs. By integrating with existing OME-Zarr viewers like ViZarr, napari and MoBIE, this allows for interactive visualisation of images and their processing results. And thanks to the federated deployment approach of Fractal, it can be hosted on a given institution’s server and integrated with local HPC to process terabytes of image data.

From Beginner to Builder: Supplier Rating App with an AI Coding Assistant

Mon, 01 Jan 0001 00:00:00 +0000

Presenter: Anusch Bachofer from EMPA

Research software is increasingly built by domain experts who are not trained software engineers. This talk shares a first-hand account of how a self-described programming beginner developed a production-oriented Supplier Rating Software (SRS) web application for a research department at EMPA, with an AI coding assistant (Claude) as a constant collaborator.

The project began as a project work in the ZHAW CAS Object-Oriented Programming course, carried out with a three-person student team, and has since matured into actively maintained software integrated with openBIS, the laboratory information system used across our labs.

The stack spans a Java 25 / Spring Boot backend, an Angular frontend, and a GitLab CI/CD and ArgoCD deployment pipeline — a breadth that would be daunting for a newcomer without support.

I will show concretely how AI assistance changed the learning and engineering process: not by writing code for me, but by explaining why, enforcing good practices (object-oriented design principles, ESLint, meaningful HTTP error handling), and letting me review and understand every change before applying it.

I will also discuss the limits and risks I encountered — over-reliance, verification, and keeping ownership of the codebase. The talk aims to give fellow RSEs, and especially researchers who code, an honest, practical perspective on where AI assistants genuinely accelerate research software development, and where human judgment remains essential.

Hospital in the cloud

Mon, 01 Jan 0001 00:00:00 +0000

Presenter: Claudia Wecker from ETH Zurich

Kantonsspital Baden (KSB) is a leading regional centre hospital serving as a primary point of care for more than 350,000 residents in the eastern region of the Canton of Aargau and the Limmat Valley. With close to 400 beds, approximately 22000 inpatient cases, 324000 outpatient cases in 2022 and a workforce of more than 2800 highly qualified employees, KSB demonstrates a firm commitment to high-quality patient care.

In an increasingly data-driven and technologically advanced healthcare environment, data represents a strategic asset of significant importance. KSB manages a broad spectrum of medical and operational information on a daily basis, including patient data, diagnoses, treatment plans, research insights, financial data, and support processes. This wealth of information holds substantial untapped potential that KSB aims to harness more effectively.

By implementing a state-of-the-art and future-proof data platform, the technological foundation required to efficiently realise high-value use cases in data management and artificial intelligence is established. To ensure full compliance with legal and regulatory requirements, all data within the platform is anonymised. By removing personally identifiable information and applying robust data protection measures, patient privacy remains fully safeguarded while meaningful analysis is still possible.

Furthermore, the platform streamlines the process of providing data for scientific and research purposes. This could facilitate stronger collaboration with academic institutions and research partners in the future, ultimately supporting medical progress and contributing to improved patient outcomes.

Keynote: Agentic AI

Mon, 01 Jan 0001 00:00:00 +0000

We are pleased to announce the following keynote by Markus Kreft from Agentic Systems Lab during the Swiss RSE day on 31 August 2026:

Research Context Engineering - Agentic Development in RSE

LLM-based coding agents have proven to be powerful tools for software development. While trend-driven approaches like “vibe coding” empower users with little to no programming experience to build surprisingly complex systems, they come with severe limitations regarding code quality, reproducibility, and long-term maintainability.

Solely relying on AI leads to a massive accumulation of “cognitive debt”, where developers gain initial speed but lose fundamental understanding of their own codebases. The emerging practice of agentic development attempts to bring the raw utility of coding agents into the rigorous, systematic workflows required by the software engineering discipline.

This talk explores how to bridge that gap in the field of Research Software Engineering. I will examine what makes the research use case fundamentally unique compared to traditional software engineering, and why guiding an AI means treating everything, from prompting to system architecture, as a context engineering problem.

Drawing on real-world heuristics, the presentation will offer pragmatic advice on when to heavily leverage agents for surgical edits or isolated projects, and when to avoid them and rely on the highly efficient expressiveness of programming languages. Finally, I will anticipate how this paradigm shift will evolve the RSE discipline, transforming daily workflows from focusing on API details and syntax to architecting robust, scientifically valid systems.

Keynote: Open Research Data

Mon, 01 Jan 0001 00:00:00 +0000

We are pleased to announce the following keynote by James Runnalls from EAWAG during the Swiss RSE day on 31 August 2026:

FAIR Enough? Why Isn’t More Data Being Reused?

A decade on from the FAIR principles, open research data is being deposited at remarkable scale. But is anyone actually using it? Too often the answer is no: the data sits in generic repositories, compliant with a checkbox version of FAIR that falls short of the original aspiration of truly reusable data.

The data that does get reused tends to live in domain-specific repositories, like the Protein Data Bank or GenBank. When a whole field deposits in one place, under shared standards and formats, the collection becomes consistent and indispensable, and gets used no matter how clunky the interface. That works brilliantly where it happens, but few fields can assemble the coordinated, sustained effort it demands, and others have data far too heterogeneous to consolidate this way.

A good open-data experience, though, can be driven either by consistency or by design. Drawing on the lessons, and missteps, of building Alplakes, a small, award-winning platform for Alpine lake data, I’ll make the case that using design to deliver an excellent user experience can provide a tenable alternative, achievable for small teams with limited resources. By making data a pleasure to access and reuse, we can bring the original vision of FAIR to a far wider range of open data.

Life Cycle Assessment Data and the Advent of Artificial Intelligence

Mon, 01 Jan 0001 00:00:00 +0000

Presenter: Michael Weinold from ecoinvent Association

This talk examines where AI-assisted life cycle assessment collides with non-negotiable ISO and academic requirements for source traceability, pedigree-based uncertainty, and transparent documentation. It walks through the concrete failure modes of a typical LLM document pipeline - table and figure extraction, chunking, embedding retrieval, and citation hallucination during synthesis - and explains why these are architectural limits of LLMs rather than bugs to be patched.

NestBot 2.0: A Controlled Multi-Agent AI System for Research Building Data at EMPA

Mon, 01 Jan 0001 00:00:00 +0000

Presenter: Michael Locher from EMPA

NestBot 2.0 is a production multi-agent AI system that gives researchers at the NEST building (EMPA Dübendorf) natural-language access to heterogeneous building data — sensor metadata, digital twin, and research outputs. This demo shows how we moved from a rigid RAG pipeline to orchestrated specialist agents: each agent has a defined tool set, runs through a central gateway with RBAC and cost tracking, and is evaluated against a structured test battery before release. No agent goes to production without a documented performance score. Intended audience: RSEs working with LLMs, research data infrastructure, or AI in science. No prior agent experience needed. Attendees will see: a live agent query against real sensor metadata, the evaluation framework that controls agent quality, and the architecture decisions that make the system reproducible and auditable.

NTSuisse: a Web Platform for LC-HRMS Data of Swiss Surface Water

Mon, 01 Jan 0001 00:00:00 +0000

Presenter: Kai-Michael Kammer from EAWAG

NTSuisse is a national Swiss water-quality monitoring platform for high-resolution mass spectrometry (HRMS) data, used by cantonal environmental agencies, water suppliers, and the Swiss Federal Office for the Environment. The platform centralises data upload, automated processing, target and suspect screening, semi-quantification, and visualisation. It runs on Azure (with a separate staging instance), while the architecture supports local deployment equally well, which is how the development team iterates day-to-day.

The platform is built and maintained by a team of three which is atypically large for a science project, but still small by any software-engineering standard. The stack is polyglot: React + TypeScript frontend, Python/Airflow orchestration, PostgreSQL + LabKey for data and auth, and Docker throughout. The backend enforces a strict layered architecture, separating pure functional logic separated from I/O, which keeps the offline test suite fast and refactoring safe. CI enforces linting, formatting, and the offline test suite at every merge request.

We also share our experience integrating AI tooling beyond code completion: project-specific instruction files, domain-knowledge skill files, and agent-callable planning workflows that give the LLM enough context to reason about architecture, not just syntax.

What you will take away:

practical patterns for structuring a science platform which is being actively used by many partners
a concrete approach to making AI assistants useful at the project level

Audience: RSEs comfortable with Python and web stacks; no mass-spec background needed.

OptiMat Alloys: a FAIR living database enabled by a conversational agent

Mon, 01 Jan 0001 00:00:00 +0000

Presenter: Yang Hu from EMPA

OptiMat Alloys reframes a data-coverage problem as a software one, and the talk is aimed at research software engineers from any domain — no materials science background needed, though some familiarity with LLM tool-calling and FAIR data helps. Instead of shipping a static dataset, it pairs a natural-language LLM agent with a simulation backend: the agent parses a request, dispatches molecular-dynamics jobs driven by foundation ML interatomic potentials, and commits every result to a shared, versioned database with full provenance, while running multiple potentials and structural realisations per composition to yield built-in uncertainty estimates.

In doing so it extends FAIR principles from pre-computed repositories to on-demand generation, built on open scientific-Python tooling. Listeners will take away a working pattern for putting an LLM agent in front of expensive scientific compute so non-experts can drive it in plain language, and how to keep machine-generated results trustworthy through provenance, versioning, and uncertainty quantification.

A public demo is available at https://youtu.be/lQzuorkzPMc

Research Software Engineering in the Age of Generative AI

Mon, 01 Jan 0001 00:00:00 +0000

Presenter: Michelle Barker from Research Software Alliance Australia

The Research Software Engineering in the Age of Generative AI workshop, held in March in Edinburgh, UK, brought together participants to explore how Generative AI may reshape the research software ecosystem, and to help inform a broader community vision for the future of the field. The workshop addressed both the opportunities and risks emerging from the increasing integration of AI into scientific workflows, including faster software development, automation, and broader accessibility, alongside concerns around reliability, reproducibility, provenance, verification, security, and workforce transformation. Through lightning talks, community discussions, surveys, and focused working groups, participants examined how RSE roles, institutional policies, collaboration models, training approaches, and software verification practices may evolve in the coming years. Discussions highlighted the continuing importance of human expertise, trustworthy software practices, and coordinated community action in an AI-enabled research ecosystem.

A major outcome of the workshop was the identification of 46 proposed pilot activities and studies spanning nine thematic areas: institutional policy and narratives, tradeoffs and risk frameworks, software attribution and incentives, verification and validation, evolving RSE roles, training and workforce development, management playbooks, equitable access to AI technologies, and new forms of collaboration between researchers and AI systems. Proposed activities range from lightweight community-driven initiatives and training communities of practice to large-scale empirical studies, ethnographic research, and longitudinal analyses of AI-assisted research software development.

This talk will give an overview of this work and progress since the workshop. See also https://www.researchsoft.org/events/rse-ai-workshop/

RésIn: local network of RSEs in Paris

Mon, 01 Jan 0001 00:00:00 +0000

Presenter: Diego Antolinos from Université de Neuchâtel

A general presentation of the project RésIn (https://projet-resin.fr/projet) and its goal (creating a RSE network between Sciences Po Paris and Université Paris-Cité). Followed by a focus on its 3 key tools: the bimonthly method seminar (canonical seminar on research methods), the monthly open workshop (a special open-doors workshop where people can get help from engineers), and the online repository of engineers’ profiles where people can sign up and display their key competences.

I intend the presentation as an inspiration for RSE Swiss from a sister-project that’s been successful-ish. And I think it can spark debate around the (in-)adaptation of some of its tools to our audience in Switzerland (e.g. the online repository is controversial).

RO-Crate Schema Plus: Self-Describing, Ontologized RO-Crate

Mon, 01 Jan 0001 00:00:00 +0000

Presenter: Andreas Meier from ETH Zurich

Main points:
Projects focused on interoperability in the ETH domain. RO-Crate Schema Plus profile is used as a common interchange format between heterogeneous systems, ELNs, repositories, with different schemas. RO-Crate Schema Plus adds a machine-readable schema with semantics to RO-Crate.

What listeners will learn:<br/ An approach to have a machine-readable RO-Crate profile with custom types.

Intended audience:<br/ People working with exchange formats, particularly RO-Crate

Storing cryogenic samples in ultra-high vacuum

Mon, 01 Jan 0001 00:00:00 +0000

Presenter: Reto Trappitsch from EPFL

To store frozen research samples after sample preparation and prior to analysis, a storage chamber that keeps them at 120 K temperature and in ultra-high vacuum (~1e-9 mbar) is required. In this talk I will discuss the development and integration of the control electronics, firmware, and host software with user-interface that now controls our home-built cryostorage solution. I will specifically point out how and why Rust is the ideal language for this development and how it can be used across the hardware stack.

Controlling the chamber requires

(a) control of certain parts via digital I/O lines,
(b) communication with several existing instruments via established protocols (e.g., via RS-232),
(c) reading, updating, and displaying the current status of the chamber and attached instruments,
and (d) interfacing with the user.

The electronics board for task (a) was designed in-house and uses a Raspberry Pi Pico 2 board as the controlling MCU. Firmware for this MCU was written in Rust using embassy. This firmware communicates with our Rust-based host software (which takes care of tasks (c) and (d)) using postcard and postcard-rpc. Finally, we also developed Rust crates to communicate with various instruments via RS-232, RS-485, and OPC UA. The host software and touch-driven user interface runs on a Seeed Studio reTerminal DM, a Raspberry Pi Compute Module 4 based Linux computer with a 10’’ touch screen. The user interface was developed using Slint.

Symbolica: a modern computer algebra system

Mon, 01 Jan 0001 00:00:00 +0000

Presenter: Ben Ruijl from Ruijl Research

This talk will introduce Symbolica (https://symbolica.io), a modern computer algebra system used by various universities that is written in Rust. I will describe what computer algebra is, why it is hard, why Rust was chosen and the importance of strong Python bindings. I will also go into the role and limitations of AI in writing complicated high-performance software.

The age of personalized software and how it might change research

Mon, 01 Jan 0001 00:00:00 +0000

Presenter: Simon Dürr from HES-SO Valais-Wallis

Showcase of the techstack I use in my work to build personalized software for my needs using AI, what I do to secure and limit exposure to security risks, some examples from industry partners (start-up, big pharma) and my takes on how this will impact the position/role of research software engineering, junior software engineers and training the next generation of RSEs

The case for research software engineers

Mon, 01 Jan 0001 00:00:00 +0000

Presenter: Erik Fröjdh from Paul Scherrer Institut

The detector group at PSI develops high-performance X-ray detectors for synchrotrons and free-electron lasers. Traditionally, software for detector characterisation and data analysis was developed by PhD students and postdocs for specific use cases, with little code sharing. However, with data rates from the next generation of detectors reaching several 100 Gbit/s, and a typical dataset from an experiment being on the order of 10 TB, this is no longer feasible.

In this talk, we show a path from a world of ‘it works on my machine,’ with custom scripts and makefiles containing hardcoded paths, to fast and reliable software. We are also improving efficiency by reducing duplicated work. As an example, we take a closer look at the calibration code for the Jungfrau detector, where, using relatively standard techniques, we were able to reduce the runtime from around 2.5 hours to 3 minutes showing the value that a Research Software Engineer can bring.

Web applications for education and behavioural experiments

Mon, 01 Jan 0001 00:00:00 +0000

Presenter: Justinas Grigaitis from UZH Department of Economics

I would like to showcase a few interactive web applications that I developed for my PhD research in behavioural economics and education. I used these applications for conducting field experiments with ~1,000 students at the University of Zurich over the past year and I am currently developing new applications for an experiment with ~600 middle-school students in India. My web applications involve AI features (e.g., chatbots and other Generative AI tools to improve student learning), behavioural tests, and surveys. I also develop separate applications for live experiment monitoring, so that PI and other researchers could see live data and always be in control without needing to run any code on their end.

The intended audience are assistant software engineers or researchers who conduct experiments with human subjects in behavioural economics, psychology, cognitive science, education, and other similar disciplines.

The main point will be to inspire ideas for new research questions that involve AI chatbots or leveraging Generative AI to facilitate the creation of behavioural experiments. In addition, I would emphasise the need for shared infrastructure and best practices within university labs or departments to make development of behavioural experiments more efficient and robust.

Finally, I would present arguments why using a modern web stack (e.g., React and Next.js) with Agentic AI is more effective than relying on third-party infrastructure like Qualtrics or oTree.

Why make it long and understandable when you can make it short and terrible?

Mon, 01 Jan 0001 00:00:00 +0000

Presenter: Samuel Gaist from Idiap Research Institute

Everybody knows that writing lots of code is bad for the shoulders, elbows and carpal tunnel, however compacting code too much will hurt the brain in the long term.

This talk will discuss coding best practices around the sweet spot of writing enough code to keep things clear and understandable without going overboard trying to be the LOC king. A science that stays true even in the age of agent coding and AI helpers.