Multi-Agent Systems for Scientific Research Hackathon

Open to all postgraduate students and research staff with an interest in using AI in research 

This hackathon will challenge participants to build multi-agent systems for scientific research workflows. Large language models are now achieving exceptional performance on coding and other scientific benchmarks. When embedded within multi-agent orchestration frameworks, these agents can tackle highly complex problems through multi-step planning, reasoning, and execution.

Participants will be invited to build on top of open-source orchestration libraries such as AG2, LangGraph, and OpenClaw, as well as multi-agent systems such as Denario or any other open-source system, in order to automate a scientific workflow of their choice.

Projects will be evaluated according to a range of criteria, including originality, technical implementation, systematic validation and evaluation, traceability, explainability, and user experience. Attendees will participate in project teams of up to 5 participants. It is possible to join the hackathon as a team or as an individual, and the organisers will support with arranging teams where needed.

This hackathon is being run jointly by the Accelerate Programme for Scientific Discovery and Infosys Cambridge AI Centre. There will be a judging panel comprised of Cambridge researchers and representatives from Infosys. Prizes will be offered to teams with the best projects as selected by the judges.

Registration