Inspiration Lab notebooks haven't changed in decades. Researchers spend too much time copying readings, formatting notes, and writing down things that should be automatic. We wanted AI to do the boring stuff so scientists can focus on actual science.
What it does LabScribble is a digital lab notebook where AI helps with documentation, not thinking. Take a photo of instrument readings and AI pulls out the values. Record voice notes and they get transcribed right away. Upload data files and get charts made for you. Ask questions and get useful answers like dilution math, protocol steps, or unit conversions.
How we built it
React 19 and Vite for the frontend with a simple timeline design. Express backend for file uploads and AI processing. Google Gemini 3 Flash handles images, audio, data files, and text questions. Deployed to Vercel with a demo mode that stores data in the browser.
Challenges we ran into
Getting the AI to stop interpreting results and just document them. Finding the right balance between user control and smooth workflow. Making chart generation work reliably with Gemini's code execution feature.
Accomplishments that we're proud of
The "caption-directed extraction" feature lets you tell AI what to look for in a photo. The accept, edit, or decline flow keeps researchers in control of their notes. We built an AI assistant that actually knows when to step back.
What we learned
AI works best as a helper, not a replacement. Researchers don't want AI drawing conclusions. They want it doing the grunt work. Good design means showing AI output before saving it, not after.
What's next for LabScribble
Team features for shared lab notebooks. Offline mode with sync. Direct connection to lab instruments. Export to standard lab notebook formats. A mobile app for use right at the bench.
Log in or sign up for Devpost to join the conversation.