Automating Research with Agents & Skills
Dr. Rahul Sundar is a PhD from IIT-Madras, currently works as an AI/ML Scientist II, and simultaneously runs Dhyuti Frontiers, a research and community initiative making AI4Science for the hard sciences more accessible to researchers. He is also closely involved with ANRF, and the Google Developer Group’s AI4Science India chapter.
In this session, he will showcase how researchers can use Skills.md files with AI agents such as Claude Code, Codex or a custom one, to automate the research workflow, covering literature survey, problem specification, code implementation, and analysis and reporting.
We will discuss takeaways on writing skill files that can be applied across domains, run through examples from his domain of expertise, Physics-Informed Neural Networks (PINNs), and explore DhyutimaanPI, a custom repo they use in-house to do automated research.
The session will end with the launch of open research problems that attendees can attempt on their own using the shared repo, or with skills written on their own.
Session Repo - https://github.com/DhyutiLABS/DhyutimaanPI/tree/main
Part of the CAISc 2026 pre-conference series. Learn more at caisc2026.github.io.
Learn more about Dhyuti Frontiers - dhyutifrontiers.info