LabMate

Inspiration

Drug discovery is slower and more expensive than ever. Labs are searching for ways to accelerate and optimize R&D. We were inspired by lab-in-the-loop robotics: automating workflow, continuously generating actionable data to feed model training and speed up lead compound identification.

What it does

LabMate is a robotic lab assistant that takes instructions from a server via a dashboard, autonomously navigates the lab, automates organization, and collects data from each task for real-time reporting. In this hackathon version, LabMate supports basic navigation and task execution.

How we built it

We developed a REST API to interface with our robot’s high-level SDK, then wrapped that with a dashboard for visualizing samples, their current protocol stage, and location in the lab. This setup enables remote instruction and tracking of lab workflow status.

Challenges we ran into

Precise arm control for real lab work is tough—much more dexterity is needed than standard SDK provides. Robot setup, movement modes, and hardware restrictions required major troubleshooting. RL policy training for stable manipulation was compute-intensive and delicate.

Accomplishments that we're proud of

Despite repeated setbacks (setup headaches, hardware quirks, and feature pivots), we managed to deliver a live demo showing server-driven instructions and navigation.

What we learned

Cool-looking robots aren’t always the best choice—lab environments demand fine motor skills and reliability. The right hardware (precise finger/joint dexterity) is critical. We learned to balance ambition and feasibility in rapid prototyping.

What’s next for LabMate

We plan to transition to robots with more dexterous hands, expand RL-powered controls for protocol automation, and design a robust data ingestion pipeline to fuel future research and rapid lab iteration.

Built With

• fastapi
• g1
• nextjs
• unitree