Case study · 2025–2026
Bridge
An ASL translation glove: flex sensors on an ESP32, streaming straight to the browser over Bluetooth.
The problem
Sign language users shouldn't need an interpreter for everyday interactions. Bridge reads hand poses through flex sensors and translates them in the browser, in real time.
Decisions, and what they cost
Every architecture is a set of trade-offs. These are the ones I made, the alternatives I rejected, and why.
Flex sensors as voltage dividers
vs. IMU gesture recognitionCheap, readable, per-finger, with a rolling average to kill jitter. IMUs capture motion better but need a training pipeline the timeline didn't allow.
BLE straight to the browser
vs. a phone app or relay serverNo install, no backend, no relay latency. The glove pairs directly to a web page. The cost is Web Bluetooth's spotty support on iOS.
Evidence it works
Working end-to-end demo: glove → BLE → live translation in the browser.
I designed the sensor calibration flow and BLE characteristic mapping.
Built with Krisha Veera, Lana Othman, Sahil Shukla, and Salim Aissaoui.
What I'd do differently
I'd replace the manual per-user calibration with a guided routine that fits finger ranges automatically, and swap the rolling average for an exponential one for faster response at the same smoothing.