Snyder Lab research
A literature tool for a Snyder Lab PhD candidate, built so she could trust what an AI pulled out of 34 papers.
- Role
- Undergraduate research assistant; built the tool
- Timeline
- Spring 2026 (Snyder Lab, Stanford Medicine)
The question
A PhD candidate studying non-invasive ways to measure gut health had 34 papers on bowel-sound analysis to work through, and the thing is that AI can summarize that much fast but also gets numbers wrong, so the real question was whether she could trust what it wrote.
What I did
I built a wiki in Claude Code where every claim links back to the page it came from and is tagged as extracted, inferred, speculative or background knowledge, so you can always tell how sure to be. When I checked five pages against the original papers, all five held up.
What came of it
I gave her the wiki along with a short brief on whether it was worth the setup, and whether to use it was her call.