Arpit Mishra
Fullstack AI Developer
Socials
APIs, databases, infrastructure
Choosing, sharding, replicating and querying data at scale
The classic 'design Facebook's feed' interview question — fan-out-on-write vs. fan-out-on-read, the hybrid systems actually run, and why ranking is a separate problem from assembly.
A classic system-design walkthrough: how to fan a single business event out to push, email, SMS, and in-app channels without spamming users or duplicating sends.
A classic systems-design interview walkthrough — real-time delivery, group fan-out, ordering, history, and presence.
Why retrying a failed request can silently charge a customer twice — and the idempotency-key pattern that makes retries safe.
Two unglamorous backend problems that quietly wreck performance under load — one about how connections are managed, one about how queries are shaped.
Two concurrent requests, one shared row, and a decrement that quietly disappears — how pessimistic locking and optimistic concurrency control each keep concurrent writes honest.
The classic system design warm-up question, worked through end to end — encoding, collisions, caching, and the 301-vs-302 trade-off everyone forgets.
A classic system-design interview question, worked through from a naive in-memory counter to a Redis-backed sliding window that survives real traffic.
Why authentication and authorization aren't the same thing, and how sessions, JWTs, and OAuth2 each solve a different piece of the puzzle.
How traffic actually gets spread across servers — round robin, least connections, and consistent hashing, plus where routing decisions get made in the stack.
Why caching is the highest-leverage performance fix in most systems — and why invalidating a cache correctly is harder than writing to one.
