Science in the Agentic Era: Structured Experimentation With Ado
summary
ado is an open-source Python framework that constrains coding agents with schema-based validation (Pydantic), provenance tracking, and a structured datastore so experiments are reproducible. It plugs agent skills for querying plans, formulating discovery spaces, running diagnostics and analysis, and uses Ray (or Kubernetes) for execution so teams run up to 2,000 concurrent optimizations. Demo on CLEX MIP tuning shows a failing run diagnosed by a single-point diagnostic and then a 200-trial campaign successfully rerun.
AI-written summary · about
chapters (9)
- 0:00Science agents and ADO
- 5:26DevOps and scientific method
- 7:13ADO core features
- 12:15Coding agents' five traps
- 13:39Agent skills for exploration
- 16:28Demo setup
- 17:11Troubleshooting failed experiments
- 20:51Agent validates new spaces
- 22:17Conclusion and ADO involvement
related talks
19mAutonomous Agents for Scientific Tasks
01h 09mMulti Agent Architecture using ADK
01h 03mSupercharge your AI agents: Inside the new ADK integrations ecosystem
25mBeyond Orchestration: An Open Source Architecture for Agentic Systems
35mThe Agentic AI Engineer
24mFrom First Experiments to AI Agents: Adoption Experience and Future Challenges
27mAgentic Coding with Discipline & Skill: Practical Techniques and Tips
40mAI Coding Agents and how to code them