From sign-up to
live cues in four steps.

No forklift upgrade, no month-long implementation. Here's the whole loop, start to finish.

Onboard your gym

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 check in

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.

Camera sessions activate

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.

Real-time cues delivered

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.

Under the hood

Built to ship fast, scale sensibly

The technical approach is deliberately pragmatic — cloud-first to reach the floor quickly, with an edge path for scale.

Lightweight pose model

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.

Cloud now, edge next

We start cloud-based to ship and iterate fast, with a clear path to edge inference on zone-mounted hardware once volume justifies it.

Rules first, ML next

Fault detection starts with joint-angle thresholds and graduates to an ML classifier once there's enough real labeled session data to train on.

Standard SaaS backend

A reliable multi-tenant stack with a background job queue for billing and dunning. This layer's job is reliability, not novelty.

See the loop on your own floor