I build systems that work offline, infer on-device, and serve live data at scale — with a hard constraint: no cloud, no server, no compromise on privacy.
My primary domain is Android development with Kotlin and Jetpack Compose, where I apply clean MVVM architecture, Hilt dependency injection, Room persistence, and Kotlin Coroutines/Flow to ship apps that feel fast and hold up under real use. Alongside that, I build TypeScript/Node.js backends — Fastify servers with two-layer Redis + LRU caching, deployed serverless on Vercel. The stack is wherever the problem is.
val developer = Developer(
name = "Sanjeevu Tarun Sree Prasad", // also: Tarun Sree Prasad
github = "github.com/Sanjeevu-Tarun",
location = "Hyderabad, Telangana, India 🇮🇳",
roles = listOf("Android Developer", "AI/ML Enthusiast", "Backend Engineer"),
stack = listOf("Kotlin", "Jetpack Compose", "TypeScript", "Node.js", "TFLite", "Python"),
education = "B.Tech CSE — Final Year · Hyderabad, India",
building = "DeviceIntel — AI smartphone analyser · BitChat — offline P2P messenger",
learning = "Advanced Coroutines, Kotlin Flow, Compose animation APIs",
philosophy = "Build offline-first. Ship production-grade. Never cut corners."
)
| 🔭 Currently Building | 📡 Live in Production | 📚 Deep-diving | 🏆 Competitive Programming | 💬 Talk to me about |
|---|---|---|---|---|
| DeviceIntel AI Analyser · BitChat | 2 APIs on Vercel with Redis caching | Kotlin Coroutines, Flow & advanced Compose | LeetCode · HackerRank | Android · Scraping · Offline systems · APIs |
"I debug at midnight and ship by dawn. Build offline-first. Make it fast. Never cut corners."
Production systems · Real data · Real users · Engineering depth
|
AI-Powered Smartphone Analyser & Comparator Runs a full TFLite recommendation engine entirely on-device — zero cloud, zero latency, full privacy. Scrapes live specs from GSMArena, persists them in Room DB, and surfaces intelligent phone comparisons via a clean Jetpack Compose UI built on strict MVVM + clean architecture. What makes it special: On-device ML inference means it works offline. The AI layer scores and ranks phones by use-case fit — not just raw specs.
|
📡 BitChatOffline Bluetooth P2P Messenger — No Internet Required Privacy-first messaging with a hard constraint: no internet, ever. Handles full Bluetooth device discovery, secure pairing, file transfer, and persistent chat history — all over raw Android Bluetooth APIs, with a Compose UI backed by Room DB and Kotlin Coroutines Flow for reactive state. What makes it special: Truly offline. Works in subway stations, remote areas, corporate environments with locked Wi-Fi. Zero server, zero data collection.
|
|
Production GSMArena + DXOMark REST API A serverless Fastify API deployed on Vercel, serving real-time smartphone specs, DXOMark camera sub-scores, and sample images. Engineered with a two-layer Redis + LRU cache for sub-100ms repeated queries. Penalty-scored smart search ranks results by relevance, not just string match. ⭐ 2 stars and growing. What makes it special: No official GSMArena/DXOMark API exists — this fills the gap with production-grade reliability and caching.
|
NotebookCheck + GSMArena + DXOMark Scraper API A 2200+ line TypeScript API that crawls and indexes NotebookCheck device reviews, SoC benchmark data, and GSMArena specs into a Redis crawl index — with a fully resumable pipeline and SearXNG search integration. Handles thousands of indexed entries with lazy loading and incremental refresh. What makes it special: The crawl pipeline is resumable — no restart from scratch on failure. Enterprise-grade reliability in a solo-built project.
|
|
Advanced Android Clipboard Manager Monitors clipboard changes in the background via Android's What makes it special: Background monitoring without a persistent notification. The categorisation engine works entirely offline.
|
🎮 GameChatReal-Time Multiplayer Gaming Platform Tic Tac Toe, Connect4, and Chess — all live-multiplayer with lobbies, spectator mode, and chat. The Socket.IO event system handles room creation, game state sync, and disconnection recovery. JWT auth with MongoDB persistence for player history and rankings. What makes it special: Full game logic + real-time sync + spectator mode, all in one solo-built full stack project.
|
Languages: Kotlin · Java · TypeScript · Python · JavaScript · C
Mobile & Frontend: Android Studio · Jetpack Compose · React · HTML · CSS · Tailwind CSS
Backend & Cloud: Node.js · Fastify · Vercel · Firebase · Redis · MySQL · SQLite
AI / ML: TensorFlow Lite · PyTorch · Google Colab · On-device Inference
Tools & Workflow: Git · GitHub · Figma · VS Code · Postman · Android Profiler
This isn't a portfolio of tutorials. These are systems built to solve real problems under real constraints:
- On-device ML inference — DeviceIntel runs TFLite models locally. No API key, no cloud bill, no privacy leak. Works fully offline.
- Offline P2P communication — BitChat uses raw Android Bluetooth APIs. Designed for zero-internet environments where privacy is non-negotiable.
- Production scraping APIs —
mobile-specs-apiandnotebookcheckerserve live data scraped from GSMArena, DXOMark, and NotebookCheck, cached via two-layer Redis + LRU, deployed serverless on Vercel. Real traffic. Real uptime requirements. - Resumable data pipelines —
notebookchecker's 2200+ line crawl engine supports checkpoint-based resumption — no data loss on failure. - AccessibilityService integration — ClipStack Pro monitors clipboard state in the background using Android's accessibility layer, without holding a persistent foreground service.
- Real-time multiplayer systems — GameChat implements a full Socket.IO event loop with room management, spectator state, and disconnection recovery.
| Capability | Implementation | Status |
|---|---|---|
| On-device AI / ML inference | TFLite · MVVM · clean architecture | ✅ DeviceIntel |
| Offline Bluetooth P2P | Android BT APIs · Coroutines Flow · Room | 🔨 BitChat |
| Production REST API with caching | Fastify · Redis · LRU · Vercel serverless | ✅ 2 Live APIs |
| Resumable crawl pipeline | TypeScript · Redis index · checkpoint logic | ✅ notebookchecker |
| Background OS integration | AccessibilityService · Room · Compose | ✅ ClipStack Pro |
| Real-time multiplayer | Socket.IO · MongoDB · JWT · React | ✅ GameChat |
Open to Android roles, API collaborations, open-source contributions, or a good engineering conversation.
→ Check the pinned repos below. Every project has a live demo or a working API.
If something looks useful, a ⭐ goes a long way. If you want to collaborate, just reach out.
Android developer from Hyderabad, India · Open to internships, freelance, and open-source work.
I go by Sanjeevu Tarun — final-year B.Tech CSE student from Hyderabad, India, building things I actually care about. Most of my projects come from a simple place: I wanted something to exist and it didn't, so I built it.
I tend to work in the parts of Android that most tutorials skip — accessibility layers, background services, Bluetooth internals, on-device ML. On the backend side, I like systems that have to stay up and handle real traffic, not just pass a demo.
Outside of code I'm into tech in general, and I like talking to people who build things. If you're working on something interesting or just want to geek out about Android or systems — feel free to reach out.