Senior Full-Stack Engineer · 5+ yrs · Python · TypeScript · React
I build systems that scale, languages that teach, and platforms that ship. From real-time exam infrastructure to a programming language in Hindi — I work at the layer most engineers never touch.
I'm a senior full-stack engineer with a bias toward building from first principles. Not because it's harder, but because it's the only way to truly own what you ship. I've written a custom interpreter, a real-time execution pipeline, and a payment-gated video delivery system — each from a blank file.
My work sits at the intersection of scale and purpose. The coding exam platform I built handles 10,000 concurrent students — not as a demo, but in production, during live exams. It also detects plagiarism and fake submissions post-exam using a two-phase pipeline: behavioural scoring followed by TF-IDF cosine similarity across the full cohort. The programming language I built runs entirely on a budget Android phone, offline, in Hindi and Marathi, because the student who needs it most doesn't have a laptop or reliable internet.
I care about systems that work under pressure and products that reach people who are usually left out. That's the thread running through everything I build.
I also contribute to open source. stringy-core — a zero-dependency JavaScript string utility library — is published on npm and built as a community contribution platform. The repo ships stub functions with clear input/output specs, inviting developers to implement and PR them. It's a small thing, but it's how I think good libraries should be built: in the open, with other engineers.
Students submit code → attempt check (Postgres) → INSERT submission → LPUSH to Redis → Celery worker BRPOP → Docker sandbox executes → result via Redis pub/sub → SSE push → browser updates. All of this in under 2 seconds per submission, at 10K concurrent users. Post-exam plagiarism runs a two-phase pipeline: behavioural scoring (paste ratio, speed anomaly, tab switches) followed by TF-IDF + cosine similarity across the full cohort.
Built for tier-3 city students aged 14–18 who have no laptop and intermittent internet.
The "zeroth step" into programming — code looks like the language you already speak.
Fully offline PWA, zero API calls, runs on a budget Android browser. The interpreter is a
published npm package with a public Compile(source, languageKeywords) API.
Learning Mode replays every evaluation step in the student's own language — no teacher required.
Adding a new language is one keyword map entry, no parser changes.
Full-stack course platform teaching deep CS fundamentals — Build a Programming Language,
JavaScript Engine Internals, Build a Real-Time System, and more. Rewritten in React 18 +
FastAPI. The standout feature: a RAG AI assistant that answers student questions grounded
strictly in course transcripts — no hallucinations. Transcripts are chunked into 180-word
sliding windows, embedded with text-embedding-3-small at startup,
then queried via cosine similarity + lexical re-ranking. Answers from GPT-4.1-mini cite
the exact lecture and timestamp, with a deep-link straight into the video player.
Payments via server-side Razorpay order creation — secret key never leaves FastAPI.
50+ pure string functions across 9 modules: case manipulation, cleaning, formatting,
masking & security (maskEmail, maskPhone, moderate), metadata extraction (URLs, emails,
IPs, hashtags, JSON), specialized operations (Levenshtein distance, balanced brackets),
and generation. Ships as ESM with both named tree-shakeable exports and a unified
_s namespace.
Built as an open-source contribution platform — stub functions with Contribution Station
comments invite community PRs.
I'm open to senior engineering roles, consulting engagements, and projects where the problem is genuinely interesting.
swanandkadam828@gmail.com