EnhanceR Symposium 2026

AI in Scientific Coding Workflows

Timetable

Event schedule

EnhanceR Symposium 2026 will take place at PSI Villigen on Monday, November 30th, 2026.

Overview schedule
9:30–10:00 Registration / Coffee
10:00–10:15 Plenary: Welcome
10:15–11:00 Keynote: Coding Agents and Software Variability: Threats and Opportunities for Replicability in Science (Mathieu Acher)
11:00–11:15 Plenary: Community Announcements
11:15–11:45 Coffee Break
11:45–12:20 Show-and-tell parallel sessions
12:20–13:00 Keynote: From Junior to Senior in the Age of AI (April Yi Wang)
13:00–14:00 Lunch
14:00–15:10 Show-and-tell parallel sessions
15:10–15:40 Coffee Break
15:40–16:25 Keynote: Quantum computing and Artificial Intelligence - where we stand and who benefits whom (Cornelius Hempel)
16:25–17:00 Show-and-tell parallel sessions
17:00–17:15 Plenary: Final Remarks & Farewell

Additionally, there is a 90-minute pre-event tour of the PSI research facilities the day before the Symposium, on Sunday, November 29th, 2026. The number of available spots is limited, so please register early if you're interested in attending.

Keynote talks

Coding Agents and Software Variability: Threats and Opportunities for Replicability in Science (Mathieu Acher)

More and more scientific fields are computational and depend on complex, evolving software that encodes hypotheses, models, simulators, datasets, and analyses. In domains such as neuroimaging, climate modelling, or even AI and software engineering themselves, there is a growing need to reproduce papers (re-running the original study to obtain the same or very similar results) and to replicate them (varying its conditions to check whether the findings still hold).
In this talk, I will first introduce the socio-technical challenges of both, based on how I teach reproducibility, from floating-point experiments to replicating research, and on two replication studies conducted without coding agents: metamorphic testing of chess engines and policy testing with MDPFuzz. Every study implicitly defines a large space of variants (code versions, configurations, datasets, environments, platforms). For scientists and engineers, and now for coding agents, the challenge is twofold. First, find at least one executable variant that recovers the original results (reproduction). Second, generate and navigate families of variants to study threats to validity, robustness, sensitivity, and generalization (replication).
I will then show the opportunities coding agents bring: they can reproduce Barbara Liskov’s 1968 PhD thesis, redo a football data analysis in a few minutes, or write a chess engine in languages humans never considered for the task (LaTeX, Brainfuck). These opportunities, however, come with responsibilities. When variants become cheap to generate, two major questions emerge, with important impacts in science: Are the variants right? And are they the right variants?

From Junior to Senior in the Age of AI (April Yi Wang)

AI coding tools are rapidly changing what it means to learn programming and grow as a software engineer. They can generate code, create tests, suggest fixes, explain errors, and answer questions on demand. These capabilities make programming more accessible and productive, but they also raise a deeper educational concern: many of the activities now being automated are precisely the activities through which novices develop expertise. Testing, debugging, decomposing problems, reading others’ code, asking questions, and learning from communities are not merely inefficient steps on the way to an answer; they are practices through which programmers build expertise. In this talk, I'll present a line of research focused on building programming tools that prioritize expertise growth over sheer output. I'll share systems that transform software testing into a peer-learning exercise and frameworks that guide learners and LLMs to break down problems together, rather than jumping straight to auto-generated code. We'll also look at how AI agents shift the balance of control for junior developers, and how chatbots can actually connect people to online coding communities instead of replacing them.

Quantum computing and Artificial Intelligence - where we stand and who benefits whom (Cornelius Hempel)

Quantum computing is often presented as the next accelerator for AI. Today the evidence points the other way: AI is already helping to build quantum computers, while the case for quantum computers helping AI remains largely theoretical. This talk takes stock of where quantum computing actually stands, from the latest error-correction records to what a fault-tolerant machine will demand of the classical computers around it. It then looks at both directions of the exchange: neural networks that decode quantum errors in real time, learning agents that keep processors calibrated while they compute, and AI agents that have begun to run quantum experiments, set against what quantum machine learning can and cannot yet deliver. Along the way it makes the case that the future of quantum computing is tied closely to high-performance computing, and to the people who build and run it.

Talks and workshops

See the contributions page for more information about the selected talks.

Full schedule
Track A (WHGA/001) Track B (WHGA/121) Track C (WHGA/021) Track D (WBBC/111) Track E (WHGA/U129)
9:30–10:00 Registration / Coffee
10:00–10:15 Welcome
10:15–11:00 Keynote: Coding Agents and Software Variability: Threats and Opportunities for Replicability in Science (Mathieu Acher)
11:00–11:15 Community Announcements
11:00–11:45 Coffee Break
Show & Tell session 1
11:45–12:20
Teaching Git, GitHub, and AI coding agents to researchers who do not come from IT
Lars Schöbitz (ETHZ)
Using AI to map complex physical state machines and improve failure recovery
Alvin Acerbo (PSI)
NTSuisse: how AI is used in the development of a web platform for Swiss surface water data
Kai Kammer (Eawag)
Where was the Soviet Ambassador to India this day, 60 years ago?
Piotr Jacobsson (UniFr)
12:20–13:00 Keynote: From Junior to Senior in the Age of AI (April Yi Wang)
13:00–14:00 Lunch
Show & Tell session 2
14:00–14:35
Agent capabilities for SQL access via the Model Context Protocol
Audrius Meskauskas (ETHZ)
Local vs. Frontier Foundation Models: Secure, Automated Data Pre-Labeling for Complex Engineering Workflows
Danuta Paraficz (Fernfachhochschule)
Beyond the Chatbot: LLM-Powered Learning for Process Engineering
Aris Marcolongo (FFHS)
AI usage patterns among researchers and RSE's, a sociological study
Michal Okoniewski (ETHZ)
Unrestricted LLMs: Research, Business and Attacker Perspectives
Ilya Misyura (FHNW)
Show & Tell session 3
14:35–15:10
An MCP server for agent-driven symbolic computation
Sudeepan Datta (PSI)
Teaching Git, GitHub, and AI coding agents to researchers who do not come from IT
Lars Schöbitz (ETHZ)
LLM in analytics of small molecules
Sasa Bjelic (PSI)
To vibe or not to vibe, that is the question
Linus Gasser (EPFL)
From Satellite Archives to Training Data: An Automated Pipeline for Multimodal Earth Observation Data Cubes
Bryant Weber-Hansen (FHNW Graduate)
15:10–15:40 Coffee Break & Networking
15:40–16:25 Keynote: Quantum computing and Artificial Intelligence - where we stand and who benefits whom (Cornelius Hempel)
Show & Tell session 4
16:25–17:00
To vibe or not to vibe, that is the question
Linus Gasser (EPFL)
Wô les moteurs - Slow down, in the time of speeding up
Dawn Duan (PSI)
From Satellite Archives to Training Data: An Automated Pipeline for Multimodal Earth Observation Data Cubes
Bryant Weber-Hansen (FHNW Graduate)
Towards Collaborative Agentic Data Science
Johannes Kirschner / Rok Roškar (ETHZ/SDSC)
Agent capabilities for SQL access via the Model Context Protocol
Audrius Meskauskas (ETHZ)
17:00–17:15 Final Remarks & Farewell
Sessions picker
Topic and host Session 1 Session 2
Teaching Git, GitHub, and AI coding agents to researchers who do not come from IT
Lars Schöbitz (ETHZ)
A1
WHGA/001
11:45–12:20
B3
WHGA/121
14:35–15:10
Using AI to map complex physical state machines and improve failure recovery
Alvin Acerbo (PSI)
B1
WHGA/121
11:45–12:20
N/A
NTSuisse: how AI is used in the development of a web platform for Swiss surface water data
Kai Kammer (Eawag)
C1
WHGA/021
11:45–12:20
N/A
Where was the Soviet Ambassador to India this day, 60 years ago?
Piotr Jacobsson (UniFr)
D1
WBBC/111
11:45–12:20
N/A
Agent capabilities for SQL access via the Model Context Protocol
Audrius Meskauskas (ETHZ)
A2
WHGA/001
14:00–14:35
E4
WHGA/U129
16:25–17:00
Local vs. Frontier Foundation Models: Secure, Automated Data Pre-Labeling for Complex Engineering Workflows
Danuta Paraficz (Fernfachhochschule)
B2
WHGA/121
14:00–14:35
N/A
Beyond the Chatbot: LLM-Powered Learning for Process Engineering
Aris Marcolongo (FFHS)
C2
WHGA/021
14:00–14:35
N/A
AI usage patterns among researchers and RSE's, a sociological study
Michal Okoniewski (ETHZ)
D2
WBBC/111
14:00–14:35
N/A
Unrestricted LLMs: Research, Business and Attacker Perspectives
Ilya Misyura (FHNW)
E2
WHGA/U129
14:00–14:35
N/A
An MCP server for agent-driven symbolic computation
Sudeepan Datta (PSI)
A3
WHGA/001
14:35–15:10
N/A
LLM in analytics of small molecules
Sasa Bjelic (PSI)
C3
WHGA/021
14:35–15:10
N/A
To vibe or not to vibe, that is the question
Linus Gasser (EPFL)
D3
WBBC/111
14:35–15:10
A4
WHGA/001
16:25–17:00
From Satellite Archives to Training Data: An Automated Pipeline for Multimodal Earth Observation Data Cubes
Bryant Weber-Hansen (FHNW Graduate)
E3
WHGA/U129
14:35–15:10
C4
WHGA/021
16:25–17:00
Wô les moteurs - Slow down, in the time of speeding up
Dawn Duan (PSI)
B4
WHGA/121
16:25–17:00
N/A
Towards Collaborative Agentic Data Science
Johannes Kirschner / Rok Roškar (ETHZ/SDSC)
D4
WBBC/111
16:25–17:00
N/A