Selected work

Work

Three projects where I went deep on on-device AI, audio ML, and computer vision.

Active

EdgeMind — On-device Voice Assistant

A privacy-first voice assistant for Android, running Gemma 4 with native audio and tool calling. No cloud, no account, no telemetry.

AndroidKotlinGemma 4 E2BLiteRT-LMNative audioTool callingOn-device AI
Active

Glosso Studio — Phoneme-Level Pronunciation Training, Offline

Offline pronunciation training that grades you on individual phonemes with an on-device wav2vec 2.0 model — not speech-to-text.

Kotlin MultiplatformJetpack ComposeONNX Runtimewav2vec 2.0RoomKoinKtor
Completed

CountIn — Vision-Based Occupancy Tracking in the Browser

Real-time people counting with computer vision that runs entirely in the browser — custom tracking and line-crossing geometry, no server.

TensorFlow.jsComputer VisionCOCO-SSDReal-Time TrackingBrowser ML