Spottr
Your gym already has cameras. Spottr turns them into an automatic workout log, with nothing for members to wear or type.
- Role
- Co-founder and CTO. I built the computer vision backend and led the front-end design and logic.
- Team
- With Felix Müller-Gliemann and Gard
- When
- Co-founder · Berkeley × Google DeepMind hackathon · Sep 2026 – present
The problem
Gyms lose about a third of their members every year, and engaged members stay about seven months longer than the rest. But logging sets on your phone between lifts is annoying, so most people don’t.
What we built
Spottr started at the Berkeley × Google DeepMind hackathon. It watches the gym’s existing security cameras, tracks each lifter’s pose, and counts sets and reps automatically. Members get a log of what they trained, and the gym gets an overview of how its floor is used.
Decisions I’m proud of
The gym pays, members use it. Members shouldn’t download anything or change how they train. That shaped everything from the business model down to the onboarding.
Privacy built in from the start. People who aren’t being tracked get blurred, and the AI model only ever sees one blurred composite per set, not raw footage.
Be honest about what’s real. In our pitch materials, real pipeline output and simulated screens are labeled differently, so nobody mistakes a demo for a product.
Where it is now
In pilot talks with a large gym chain. The code is private, so get in touch if you’d like a walkthrough.

Spottr replay with a skeleton overlay on a recorded squat

The gym dashboard. Who is in right now, how full the floor is and tracked sets this week (demo data)

The members view. Visits per week, a retention curve and the members who are slipping (demo data)

The member app. Reps, sets and the muscles worked, logged by the gym camera from a real clip