I build backends — specifically the parts that have to stay correct once things go concurrent. Event-driven services, message queues, caches, and the design work underneath them: where state lives, what the interfaces promise, and what happens when a node comes back after being gone for ten minutes.

Three things have most of my attention right now.

Distributed systems. I'm building a collaboration platform on CRDTs and a Kafka event bus, where the interesting work is the reconnect path and the conflict resolution, not the happy path.

System design, especially low-level. I work through LLD problems in the open and write down the reasoning — why this abstraction, what breaks at scale — rather than just committing the class diagram. It's the part of engineering I find genuinely fun.

Infrastructure. Publish-and-distribute pipelines, storage layers, versioning, the plumbing that everything else assumes works.

Most recently I was a backend developer intern at OpsAI, building the APIs and governance pipelines between AI agents and enterprise systems — role-based approvals, policy checks, risk classification, audit trails. Compliance code is unglamorous and completely unforgiving, which turned out to be good training.

Competitive programming is where the instinct came from. 1000+ problems solved, LeetCode Knight at 1880+. It taught me to go looking for the input that breaks something before I go looking for the one that works — which is most of what backend engineering turns out to be.

Studying Electrical & Computer Engineering at Jamia Millia Islamia and BS Data Science at IIT Madras, concurrently, since 2024.

→ Fastest way to reach me: sarkarsohan.3706@gmail.com

_{Last reviewed August 2026.}

npm-package-upload-infra — Publish-and-distribute infrastructure for npm packages: the upload path, versioning, and the storage layer behind it. _{Node.js · AWS}

Low-Level-Design — Working through LLD interview problems in the open — parking lot, rate limiter, elevator, and the rest — with the design reasoning written down, not just the code. Updated most weeks. _{Java/C++ · design patterns}

An event-driven collaboration workspace built for engineering teams specifically, not a Notion clone. The thesis is that everything should be connected: a document knows about the pull requests, incidents, ADRs and people attached to it, so you can ask why an architectural decision was made and trace the answer.

Deliberately built in slices rather than all at once. Repo goes public when the collaborative editing layer survives a reconnect storm.

▌ More telemetry — the year as a city, the year as lunch, per-language breakdowns

_{Ordered by what I'd reach for first, not alphabetically.}

Backend	Also Fastify · REST API design · async processing
Distributed & data	Also OpenSearch · CRDTs · event-driven architecture · distributed locks
Cloud & infra	S3 · CloudFront · CI/CD
Languages	C++ / Python for contests, Java for LLD, JS/TS for everything else
Design	Low-level design · design patterns · system design · API contracts
Frontend secondary	GSAP · WebGL shaders · Locomotive Scroll — motion-heavy client work
ML coursework	LSTM classification · YOLO object detection

I'm looking for backend and distributed-systems work and I'm available now. If you're building something where correctness under concurrency actually matters — event pipelines, real-time state, anything with a hard consistency story — I'd like to hear about it.

_{"The bug is in the case you didn't write a test for."}