← Projects
Experiment
MetriaLab LIMS
A lab management system that uses language models for the boring part: transcribing, validating and organising results.
- Category
- Software for laboratories
- Stack
- Next.js · Python · Claude API · AWS
- Site
- Visit
Labs generate an absurd amount of data and move it by hand between spreadsheets, emails and Word reports. Human error isn't a theoretical risk there: it's the day to day.
This experiment tests a simple idea — that a language model can take over the tedious part (reading, transcribing, cross-checking, flagging what doesn't add up) and leave the technical judgement where it belongs: with a person.
Why it's an experiment
Because I still don't know whether the accuracy is good enough for a context where being wrong has real consequences. I'm measuring that before promising anything.