Wearable Distributed Fitness Biomechanics Sensing

Fitness wearables measured biometrics: heart rate, calories, steps. None of them could tell you whether you were doing the exercise right. Form is what a personal trainer watches, and form is biomechanics, not biometrics.

A scalable fitness and wellness platform built on a network of small wearable sensors, each with an IMU, that captures movement quality across the body. State-of-the-art ML and DSP turn the streams into exercise identification, rep counting, range-of-motion scoring and specific coaching to correct biomechanical dysfunctions while the exercise is happening. Portable, versatile across sports, and aimed first at gym and fitness enthusiasts, with the goal of being the first wearable that acts as a personal trainer.

I led a cross-functional team of designers, backend engineers and AI/ML engineers. I also personally developed the front end of two iPhone applications in Swift: an internal data collection app used for months by the team to gather labeled IMU data for training, and the user-facing prototype app that shows the trained system identifying exercises, counting reps, managing recovery time and coaching form from live sensor data.

Wearable nodes stream IMU data to the phone. ML models identify the exercise from the motion signature; DSP algorithms detect activity, segment motion, count repetitions and evaluate range of motion to produce a movement quality score. After each set the user sees reps completed, a quality score and personalized feedback. Early prototyping used the Nordic Thingy:53, a sensor-rich development device with wireless connectivity, as a stand-in for the final hardware.

A working platform prototype with trained models, two iOS applications and a concept film demonstrating real-time coaching. The project fed LG's Edge AI work on running intelligent systems reliably on resource-constrained devices.