▤ Mobile
Kotlin, Java, Android SDK, Jetpack (Compose, ViewModel, Room, WorkManager, Navigation), MVVM / Clean Architecture, Coroutines, RxJava & RxKotlin, Flutter & Dart, Material Design, NDK (FFmpeg, Mp4Parser), offline-first sync.
Senior Mobile & AI Engineer with 12+ years shipping Android and Flutter products for security, marketplace and field-service platforms — and now building LLM-powered features in Python: retrieval, agents and the harness that keeps them honest.
I started in 2013 building small Android utilities in Delhi, then spent three years leading a 2–4 developer pod at Shineweb shipping social and field-service apps — live video, on-device video compression with FFmpeg, sockets, maps. Since 2018 I've been delivering client and product work at scale — currently on a large consumer security and smart-home Android platform, where the product details sit under NDA.
My default is full ownership: architecture, the last 10% of polish, release engineering, and the post-launch stability work nobody tweets about. I care about crash rates, cold start, offline behaviour and whether the thing still makes sense six months later.
Lately I build with LLMs. Not demos — grounded retrieval, tool-using agents, an orchestration harness with retries and permissions, and evaluation that tells me whether a prompt change actually helped. Python and FastAPI on the server, Kotlin or Flutter on the client.
Kotlin, Java, Android SDK, Jetpack (Compose, ViewModel, Room, WorkManager, Navigation), MVVM / Clean Architecture, Coroutines, RxJava & RxKotlin, Flutter & Dart, Material Design, NDK (FFmpeg, Mp4Parser), offline-first sync.
Python, LLM app development (OpenAI, Anthropic Claude, Gemini), prompt & context engineering, RAG with embeddings + pgvector/FAISS, LangChain / LlamaIndex, tool use & function calling, agent harness & orchestration, MCP-style tool servers, evaluation, guardrails, token/cost control, FastAPI.
Node.js, Express, REST & JSON APIs, WebSockets, React, Vue, Nuxt, JavaScript (ES6+), HTML5 / CSS3, Firebase, AWS Amplify, JSP / Servlets, Nginx / Apache.
MySQL, SQLite / Room, PostgreSQL, Oracle, MongoDB, Firebase Realtime DB & Firestore, Redis, vector stores, caching layers.
Git / GitHub, Android Studio, Xcode, VS Code, Postman, CI/CD, Play Console & App Store Connect, Agile / Scrum, code review, unit & UI testing, AppSec basics.
Android (5 → 15), iOS, Linux / Unix, Windows, macOS, Flutter desktop.
Consumer security & smart-home Android platform · details under NDA
IT services & mobile solutions
Product studio · led 2–4 developer pods
Android product house · consumer utility and social apps
Flagship store products plus the client builds that shaped how I work. Every link is live.
Distribution platform · Android + web
Shopkeepers issue QR code strips so mechanics redeem product discounts at the counter. Native Android app plus a Nuxt web admin, both on one Node.js API.
Project management · mobile + desktop
Agile project management for distributed teams: roadmaps, sprint tracking and real-time release reporting, from a single Flutter codebase across phones, tablets and desktop.
Pet-lover social platform: profiles, rescue and sitting services, adoption, lost-pet alerts and rich media sharing.
Find nearby parties in real time — create events, sell tickets, watch live video previews and live comments before booking.
Portfolio and rating app for the construction trades — before, during and after galleries shared with clients in seconds.
Network for founders: customisable profile pages, built-in messenger, nearby event discovery and an industry news feed.
Medical products marketplace with separate customer, seller and admin portals.
Parents find tutors, instructors accept jobs — registration, job posting and applications, on React + Firebase.
Device utilities: back up installed APKs to storage, plus an interactive water-ripple live wallpaper using your own photos.
Offline Hindi→English dictionary with pronunciations via text-to-speech, built for low-connectivity users.
The interesting part isn't the prompt — it's retrieval quality, tool boundaries, retries, cost and knowing whether the last change made things better.
Document-grounded Q&A over internal knowledge bases. Hybrid chunking, embedding + keyword retrieval with reranking, citation-linked answers and streaming responses.
Eval harness tracking faithfulness and answer relevance on every release.
Orchestration layer for multi-step, tool-using agents: pluggable tool registry, structured function-calling loop, retries, timeouts and permission scopes.
Step-level evals with latency and cost budgets per workflow.
LLM assistant embedded in a mobile product — streaming chat, prompt templates, tool calls into app data and graceful offline fallback.
Token budgets and caching to keep per-session cost predictable.
Open to senior / lead mobile engineering roles with AI scope, and to contract work on Android, Flutter or LLM product builds. Usually replies within a day.