Android · Flutter · on-device AI.
I turn ambiguous ideas into tested mobile products on Google Play — including private on-device LLM apps — then own architecture, verification, and release.
Portfolio · LinkedIn · Google Play · Resume · Email
| Product | What I shipped | Evidence |
|---|---|---|
| Chakuli | Kotlin + Jetpack Compose assistant with on-device LLM inference (LiteRT-LM / GGUF) | Play · Portfolio |
| PyMaster | Flutter Python-learning product, offline-first, billing, Crashlytics | Play listing |
| Keepary | Flutter document scanner: on-device OCR, local PDF tools; Play closed testing | Play / Devanshu Studios |
Commercial app source is private. Public engineering is below.
| Repo | Why it is here |
|---|---|
| aegis-edge | On-device Gemma 4 disaster triage on Android (weekend, inspectable) |
| Anvil | Shipped multi-service PDF pipeline |
| Scribe-Android | Kotlin contributions in an established Wikimedia project |
| HYDRA-shrinkflation-watch- | Live collectors + self-healing extraction |
| bhanu-dev82.github.io | Portfolio |
| rote-play-artifacts | Measured agent-procedure study (supporting, not the headline) |
AI-native means I pick models and coding agents by task, then keep architecture, tests, privacy, and Play-release quality. The model is output. I own the product.
- Open to remote or on-site product, Android, Flutter, and on-device AI roles.
- Training a compact on-device model for Chakuli (product work, not a separate ML identity).
- Upstream contributions to Gemma / Gemini Cookbook / LiteRT land when a PR is merged. Clones are study workspaces, not portfolio pieces.
Skills: Kotlin, Jetpack Compose, Android, Flutter/Dart, on-device LLMs, LiteRT-LM, ML Kit, MediaPipe, Firebase, Python, CI/CD.