Backend engineer in Bangalore — Java, Spring Boot, Kafka, Postgres.
By day I work on Vegapay's credit platform: I own the Credit Card Management System and the REST/SOA layers on a multi-tenant stack doing 10K+ TPS and 100M+ transactions a month, and I built its reporting service and SFTP → Kafka → S3 ingestion pipeline end to end.
By evening I build the thing underneath the abstraction, to find out whether I actually understand it. Everything below is written from scratch — no framework doing the hard part.
Four backends I built and deployed. They're live, and the point is that you can break them:
| What it proves | ||
|---|---|---|
| FrontRow | 1,000 concurrent buyers, 3 seats, 0 oversold — a Redis hold in front of a conditional Postgres write, plus a transactional outbox | how it works |
| Raft lab | Five-node consensus you can kill, partition and delay — then watch re-elect and heal | code · how it works |
| Liveboard | A multiplayer canvas that converges without a central lock — grow-only-set CRDT, Lamport ordering | how it works |
| LLM gateway | Keeps serving when a provider falls over — routing, cache, SSE, circuit breaker | how it works |
- raft-from-scratch — Raft from the paper in plain Java: leader election, log replication, the §5.4 safety rules, over a transport built to be broken. 9 tests, including a 3∥2 partition that proves the minority commits nothing.
- Kafka-cpp — a Kafka broker in C++: wire protocol, partitions, consumer offsets.
- MyOwn-SQLite-in-cpp — SQLite's file format, B-tree pages and a query executor.
- Myown-Git-in-cpp — blobs, trees, commits, the object store.
- MyOwn-DNS-Server — message parsing, record types, recursive resolution.
- Myown-grep-in-java — including a hand-written backtracking regex engine.
- My-own-working-HTTP-server-in-Java — request parsing, concurrency, static serving.
Before backend I spent a while on quantitative research — 35+ predictive signals across US and China equities, top 0.5% globally (rank 26) in AlphaVerse Premier League. That work is why I default to asking whether a result holds out-of-sample; "looks great in dev" is the same shape as "looks great in-sample." I also published NLP research on entity-based privacy for legal documents (IJSRCSEIT, 2023), and qualified for ICPC at rank 205.
namangupta.me · resume · LinkedIn · namangupta010904@gmail.com
Open to backend / distributed-systems roles (SDE1–SDE2) — Bangalore, Hyderabad, Pune, NCR, or remote.