backend · distributed systems · 3rd yr cs at kj somaiya, mumbai
I like the parts of software you only notice when they break. So I build them myself. A rate limiter that held 24.7k req/s, a write ahead log that shrugs off crashes, a scraping pipeline that runs around the clock.
Third year CS at KJ Somaiya, Mumbai. Most people meet software at the UI. I kept wondering what was underneath it. What actually happens when a database says “saved”? How does a server say “no” to the 10,001st request without falling over?
So I stopped just using infrastructure and started rebuilding it myself. A write ahead log in C++. A rate limiter in Go. Bloom filters. Quadtree compression. No tutorial told me to, I just wanted to see the guts.
A rate limiter that guards APIs like a bouncer with a perfect memory. Token buckets in Redis, drained atomically by Lua. Redis dies? A circuit breaker flips to in memory buckets and it keeps serving. Survived 3.08M requests in stress tests.
Write ahead logging built from nothing. Same durability trick that lives inside Postgres, SQLite and Kafka. Every record gets a 16 byte header and a CRC32 checksum. Kill the process mid write, restart it, and it replays up to the last valid record.
A scraping pipeline that watches flight prices across 3 airlines all day. A scheduler feeds BullMQ queues, workers scrape and clean the data, everything lands in Postgres. Bookmark a flight and it pings you the moment the price drops.
Lossy image compression in C++17. Recursively splits the image where the detail lives and flattens where it doesn’t. Took a 15.12MB image down to 661KB and you can still tell it’s a cat.
I own the backup pipeline that copies every Supabase table into SharePoint, built on Deno edge functions and the Microsoft Graph API. Getting it reliable meant paginating reads so large tables never time out, using a service role key to work around RLS, and moving from XLSX to CSV when the writer kept hitting edge memory limits. Outside that I work across the backend, indexing slow queries, writing edge functions that send messages, and cleaning up the UI where it needs it.
Backend work on a health consultation platform doing real volume. Node and Express APIs taking 10k+ requests a day, WebRTC video powering 1:1 consultations at 2,000+ sessions a week, and REST endpoints feeding the React frontend.
Joined KJ Somaiya for CS. Had never written real code before this.
Wanted to know how Postgres saves data without losing it. That question ate most of my semester.
Backend intern at CareSaathi, a healthcare startup in Pune. Within a month my endpoints were handling 10k requests a day.
Built a WAL in C++ and a rate limiter in Go to check if I actually understood the theory. Load tested the limiter to 3 million requests.
SDE intern at ZenUp Health. Reading DDIA whenever I get time.
The tools I actually reach for. Not a wishlist.
gogoroutines · channels · syncUP ~1.5 YRS
c++17memory · RAII · bit tricksUP ~2 YRS
node.jsexpress · bullmqUP ~2.5 YRS
redislua scripting · atomic opsUP ~1 YR
postgresschema design · profilingUP ~2 YRS
dockerimages · composeUP ~1 YR
prometheus + grafanadashboards · load tests · p99 trackingUP ~1 YRLooking for backend / systems internships and interesting collabs.
My inbox has zero rate limiting.