SAFESTRIDE
“Fall detection at the edge. No cloud. No latency. No compromise.”
Team Leader · Hardware + Software
Active
Overview
SafeStride is a wrist-worn edge-AI fall detection wearable I built for elderly users. It uses onboard IMU sensor data and a TensorFlow Lite Micro model to detect falls in real-time — entirely on-device, with zero cloud dependency and sub-5ms inference latency.
Built under team name GuardianPulse (Team ID: N25C01A0047), this represents my deepest technical work — bringing together embedded systems, ML model optimization, and product design into a single wrist-worn device small enough to forget you're wearing it.
Screenshots
Key Features
Tech Deep Dive
The Arduino Nano 33 BLE Sense Rev2 gives me a 9-axis IMU and BLE radio in a compact form factor. I trained the TFLite Micro model on a labeled dataset of fall vs non-fall IMU sequences, then quantized it to INT8 to fit within the MCU's 256KB flash. My C/C++ firmware handles sensor polling at 100Hz, feeds the inference engine, and triggers BLE broadcasts on fall detection.