Hi, I’m

Priyanshu Sinha.

I build the backends that keep products fast — and the AI that runs on top of them.

Backend engineer at StudyIQ, where I maintain and am responsible for 20+ Java/Spring Boot microservices within a 35+ microservices infrastructure, serving traffic peaks of 100K+ requests a minute. Outside work, I build AI agents, web, mobile and backend products for clients.

NOW
SDE-1 Backend · StudyIQ
BASED IN
Gurugram, India
FOCUS
Distributed systems + AI
OPEN TO
Freelance projects
See my backend in action ↓Work with meGitHubLinkedInRésumé

Backend + AI, live

Ask me anything. Then watch your question travel through my backend.

Every AI answer is a backend problem: rate limits, caches, retrieval, streaming. Send a question and see each hop. Then ask the same one again.

LIVE · POST /api/ask
requests: 0cache hits: 0tokens left: 5/5
01
API gateway
auth + routing
2 ms
02
Rate limiter
Redis token bucket
1 ms
03
Cache
Redis lookup
1 ms
04
Retriever
vector search, top-3
38 ms
05
LLM
generate answer
840 ms
06
Stream
SSE to browser
streamed
⏳Slow-mo is enabled by default so you can watch each hop. Want real speed?
HTTP/1.1 —
X-Cache: —
X-RateLimit-Remaining: 5
X-Response-Time: —
X-Playback-Mode: SLOW-MO · Educational pacing active (click button to benchmark real speed)
Pick a question and send it.

Try it: send a question, then send the same one again — the cache answers it without touching the model. Spam it and the rate limiter steps in. Stages are slowed down so you can see them; the labels show real timings. Switch to real speed to see how fast it actually is.

For developers

Prefer the terminal? So do I.

My whole profile is an API. Fetch it, pipe it through jq, or hire me straight from the command line.

$ curl https://api.priyanshu.dev/me { "name": "Priyanshu Sinha", "role": "Backend Engineer (SDE-1)", "stack": ["Java", "Spring Boot", "Kafka", "Redis", "AI Agents", "RAG / pgvector"], "openTo": "freelance", "hireMe": "POST https://api.priyanshu.dev/hire-me" }
01 / The LabBackend concepts, running liveRate limiting, caching, locks, retries — each with a diagram, a live demo and the code.02 / AIHow I design AI-powered systemsRAG, LLM gateways and voice pipelines, explained from the backend up.03 / JourneyThe person behind the codeEvery phase so far — the wins, the hard years and what they taught me.

Inside the Lab

Start with rate limiting →
Rate limitingSpam a button, watch the token bucket drain and the 429s arrive.LIVEDatabasesQuery plans and indexes, the N+1 problem, and a race condition that pays out twice.LIVECachingSame request twice — see the latency drop and X-Cache flip to HIT.LIVEFlash sale without oversellingHundreds of buyers, ten items, one distributed lock.LIVEIdempotent paymentsDouble-click “Pay” and still get charged once.LIVECircuit breakerBreak a downstream service and watch it fail fast, then recover.LIVEQueues, retries and dead lettersMessages fail, back off, retry, and land safely in a DLQ.LIVE

About

I started as a backend intern on a U.S. healthcare platform and now run production services at StudyIQ. On the side I take on freelance projects — web, mobile and the backends behind them.

I care about the parts of software nobody sees until they break: queues, caches, transactions, and what happens at peak traffic. I build AI-powered features the same way — backend first.

Download résumé (PDF)