Venture
RepSense AI
Train to your true capacity - not a guessed RPE.
The problem
Most lifters estimate reps in reserve by feel. That guess drifts under fatigue, caffeine, sleep debt, and ego - so progress stalls or form breaks down.
Our approach
We stream set-level heart rate from wearables, extract response features (rise rate, peak, recovery slope), and map them through a calibrated model to estimated reps in reserve. Those estimates feed session guidance - when to add load, hold, or stop - so training decisions track physiology instead of subjective RPE.
Overview
RepSense AI turns wearable heart rate into a proximity-to-failure signal. The pipeline ingests per-set HR time series, normalizes for individual baselines, and predicts remaining quality reps before form or recovery collapses. Lifters get concrete cues mid- and post-set, which cuts junk volume and makes progressive overload easier to trust week to week.
Capabilities
What it does
Set-level HR feature extraction (rise, peak, recovery)
Per-user calibration so estimates stay personal
RIR predictions wired into load and volume guidance
Works with common wearable streams and training logs
Stack