No forklift upgrade, no month-long implementation. Here's the whole loop, start to finish.
We migrate your members, subscriptions and payment tokens from your current tool. Founder-led onboarding for pilot gyms means someone's with you the whole way — no ticket queue.
Members arrive and check in through the app, a front-desk kiosk, or your existing access-control system. Every visit becomes a signal in the data layer.
At a coached station, a phone or zone-mounted camera picks up the set. The pose model locks onto joints and starts tracking reps, tempo and form in real time.
The moment form drifts, the member gets an audio or on-screen cue — mid-set, not after. The session then logs back to their profile and feeds the churn-risk model.
The technical approach is deliberately pragmatic — cloud-first to reach the floor quickly, with an edge path for scale.
A MoveNet/BlazePose-class architecture fine-tuned on gym footage — usable frame rates on modest hardware, because a camera setup isn't a data center.
We start cloud-based to ship and iterate fast, with a clear path to edge inference on zone-mounted hardware once volume justifies it.
Fault detection starts with joint-angle thresholds and graduates to an ML classifier once there's enough real labeled session data to train on.
A reliable multi-tenant stack with a background job queue for billing and dunning. This layer's job is reliability, not novelty.