ProjectsJul 2025 to Nov 2025
Project Talaria
A shoe that tracks heart rate and gait, and forecasts both.
- Embedded systems (ESP32)
- Sensor integration
- Recurrent neural networks
- Time-series forecasting
- Firebase
Impact
- R², forecasting 15 signals
- 0.97
- steps ahead
- 50
- training sequences
- 50,000+
A working wearable, streaming live, with forecasts that closely follow the real signals.
What I did
- Built a wearable on an ESP32 that streams heart rate, blood oxygen and motion-based gait data in real time.
- Sent every reading to a cloud pipeline for continuous monitoring, with live charts on the web.
- Trained a recurrent neural network on 50,000+ sequences to forecast 15 features 50 steps ahead, with an R² of 0.97 and a mean absolute error of 0.103.
The aim
Heart and gait data usually means a visit to a clinic. Talaria puts the sensors in a shoe, so they can be watched all day, and forecasts where they are heading.