CLARIPHY — AI agents and Large Language Models for Scientific Discovery
This hands-on tutorial introduces AI for scientific research, covering LLMs, multimodal data, RAG, and AI agents, using NRP resources. It culminates in JFC, a multi-agent framework demonstrating end-to-end HEP analysis workflows
Facilitators: Daniel Diaz and Andrzej Novak (MIT)
Schedule
| Start | Lesson | Duration |
|---|---|---|
| 00:00 | Setup | 0 min |
| 00:00 | Introduction — LLMs on NRP | 20 min |
| 00:20 | Launch the Agent — JFC Physics Analysis on NRP | 15 min |
| 00:35 | Chat with LLMs — Python, Multimodal, Embeddings & RAG | 50 min |
| 01:25 | Agentic Workflows — opencode & IDE Integration | 45 min |
| 02:10 | Build a Simple Agent — Tool Calling & the Agent Loop | 40 min |
| 02:50 | Check In on the Agent — How JFC Ran, and What It Found | 10 min |
| 03:00 | Finish |
Times are cumulative and assume a prompt start.
Code and other resources used for this training session can be found in the materials/clariphy branch.