Automating scientific discovery with reasoning agents
Andrew White
FutureHouse
GTC DC Biopharma Roundtable
October 2025
FutureHouse Structure
- Non-profit founded in 2023
- Funded primarily by Eric Schmidt
- Based in San Francisco
- 25 employees
Science is changing independent of AI
Arxiv.org,10.6084/m9.figshare.17064419.v3
FutureHouse Mission
Accelerate Scientific Discovery with Agents
What is an agent?
Agent: trained, makes decisions
Environment: untrained, has tools, state
Literature agent
- Literature agent that can download and read through papers
- Clinical trials, patents, OpenTargets, Metadata, citation graph
- Input: "Has anyone used LLM agents to automate the discovery of a novel molecular chromophore?
Aviary: training scientific agents
Siddharth Narayanan, James D. Braza, Ryan-Rhys Griffiths, Manu Ponnapati, Albert Bou, Jon Laurent, Ori Kabeli, Geemi Wellawatte, Sam Cox, Samuel G. Rodriques, Andrew D. White
arXiv:2412.21154 [cs.AI], https://doi.org/10.48550/arXiv.2412.21154
Training Agents
Training curves
With NVIDIA: training reasoning agents with RL
- Monthly Active Users (June) >70,000
- Tasks per minute: >1,0000 per minute
- Research Papers 150,000,000
- Wiki page for all diseases every 2 hours
- Check for contradictions 11M papers / year
- All Wikipedia every 6 days
Complete cycle of disease to mechanism to target to drug
ROBIN: A Multi-Agent System for Automating Scientific Discovery
Ali Essam Ghareeb*, Benjamin Chang*, Ludovico Mitchener, Angela Yiu, Caralyn J. Szostkiewicz, Jon M. Laurent, Muhammed T. Razzak, Andrew D. White†, Michaela M. Hinks‡, Samuel G. Rodriques