Hands-Free Lab Assistant Powered by Edge AI

Last night my team placed 2nd in the ExecuTorch Edge AI hackathon GitHub HQ. We built Beaker, a mobile application that guides and documents experiments in wet labs. Scientists doing life-saving research need to read dense protocol instructions and take lots of notes while working. Both of these activities significantly delay experimentation and discovery. Scientists need a hands-free, low-latency, and private interface for navigating and documenting protocols. This is where edge AI shines. Beaker uses native LLM, STT and TTS to make this possible, so scientists can spend more time doing science.

If you like AI/ML engineers who can execute, check out my teammates: Sanskar Thapa, Jai Dhiman, Jaden Rodriguez, Brayckner Bueres Torres.

Huge thanks to my friend Lucas Mair, who spent an hour on the phone with me explaining the pain points of executing protocols in a wet lab, and why more automation is needed. His startup ProviGenAI is is working towards a future where bioengineering is as easy as vibecoding.

We were able to build an edge AI application in just a day thanks to the incredible work being done in this space by PyTorch, Qualcomm, Meta, and Samsung Electronics. Each team was given a Samsung Galaxy S25 equipped with Qualcomm’s Snapdragon 8 Elite processor. Using GitHub Copilot, it took about 30 minutes to get a React Native application deployed to our device running Meta’s 1B parameter Llama model. It’s incredible how far edge AI has come, I’m excited to continue building in this space.