Remote-first, filing from Kolkata — Anirban Das is a full-stack developer who builds the boring-but-critical parts well: auth, data, real-time, and the AI features layered on top. He started chasing clean UIs; now he chases correctness under load.

Four years and four teams: document intelligence at Bau.ai in Berlin, AI-assisted diagnostics at Medecro.ai, an in-browser OS frontend at ProjectX, and legal-tech at Legabyte. The through-line never changes: own the data model, the API, and the frontend — then stay through the on-call.

• An Android Home Screen Where the App Name Is the Icon — A text-only Android launcher, live on Google Play — app names set in type instead of icons, every swipe reassignable, and a widget stack driven by broadcasts rather than polling.
• A Terminal Typing Trainer That Ships on Homebrew — Monkeytype, but it lives in your terminal — 15+ themes, a 25-lesson touch-typing curriculum, and an error heatmap, rendered with React's Ink and shipped via Homebrew and npm.
• A GitHub Analytics Dashboard for the Numbers-Curious — Point it at any username and it draws the picture: contribution heatmaps, language charts, and a composite profile score — all cached hard with TanStack Query.
• A Privacy-First ATS Resume Analyzer — Upload a resume, get an ATS-style report across five dimensions and 170+ skills. The engine runs entirely in memory — files never touch a disk — wired end-to-end with tRPC.
• A Notion-Style Blogging Platform with AI On Tap — A blogging platform that feels like Notion and writes like a co-author: slash commands, AI generation via the Vercel AI SDK, Better Auth, and Drizzle over Postgres.
• A Real-Time Movie Database, State Managed — Built to get the fundamentals right: Redux for predictable global state, real-time fetching, and error handling that fails gracefully instead of white-screening.

• Full-stack, end-to-end. From schema to pixel. I design the data model, build the API, wire the frontend, and stay through the on-call. PMs get estimates that hold; users get features that don't flake.
• AI features that actually ship. X-ray analysis, transcription, document auto-fill, RAG. I make the model useful inside a real product — with guardrails, evals, and a sane fallback for when it's confidently wrong.
• The un-glamorous core. RBAC, multi-tenancy, WebSockets, PDF pipelines, migrations. The parts nobody demos but everybody depends on at 2 a.m. My natural habitat.

anirbandashere@gmail.com · anirbandas.com · LinkedIn · X · DEV · Medium