Sandra

Inspiration

Bioinformatics research is bottlenecked by tooling, not ideas. A scientist with a clear question still spends days wiring together folding models, sequence analysis, and data pipelines before seeing a single result. We wanted to collapse that gap.

What it does

Sandra is an AI agent for bioinformatics. You describe an experiment in plain language, and Sandra orchestrates the full pipeline—protein folding, sequence analysis, and downstream processing—returning results in minutes instead of weeks.

How we built it

Sandra plans a research workflow from a natural-language prompt, dispatches each step to the right tool, and chains outputs into a coherent result.

Challenges we ran into

Folding model access and compute were the main constraints—orchestrating long-running jobs without blocking the agent took real engineering. Reliable plan generation across varied research questions was also non-trivial.

Accomplishments that we're proud of

We built a working end-to-end agent that takes a research question and returns real results, with no manual pipeline wiring in between.

What we learned

How to structure agent workflows around long-running scientific compute, and where LLM planning breaks down on domain-specific tasks along with fine tuneing open source models.

What's next for Sandra

Broader tool coverage, support for more research modalities, and tighter integration with lab workflows so Sandra fits into real experimental pipelines.

Built With

• anthropic
• claude
• devin
• next.js
• pi-coding-agent-sdk
• react
• supabase-(postgres-+-pgvector)
• tailwind-css
• typescript
• zod