VerdeLab Research Agent

• 

  Radiogenomic analysis of 176 KIRC patients testing whether imaging modality (MR-present vs CT-only) correlates with tumor molecular subtype.

• 

  Statistical analysis of imaging modality across 32 TCGA cancer types.

• 

  Conclusion that imaging modality is not associated with molecular subtype in ccRCC, derived from enterprise data accessed through Craft.

Inspiration

There is not a research agent for non-technical biologists in the market right now. Researchers still rely heavily on fragmented chatbot-based workflows that lack end-to-end orchestration, traceability, and automation.

What it does

Full-cycle research workflow orchestration, tool execution, and traceable research auditing.

How we built it

We feed broad pain points collected from biological researchers to coding agents, and ask them to build a platform that can be easily managed by non-technical users.

Challenges we ran into

We've been exploring a lot to provide intuitive UX designs that help users understand and monitor complex agent reasoning and iterative execution processes.

Accomplishments that we're proud of

We've developed a MVP that empowers non-technical biologists to manage the full research lifecycle—from hypothesis development and data analysis to execution tracking and audit.

What we learned

Auditability of AI research workflows is essential for building trust in life science discoveries, and for providing transparent records of data and decisions.

What's next for VerdeLab

We are expanding VerdeLab’s capabilities by integrating enterprise biological data sources through protocols like MCP to enable secure, seamless access to organizational knowledge and accelerate AI-driven research workflows.

Built With

• craft
• langgraph
• nebius
• neon
• next.js
• snowflake
• typescript
• vercel