SDE-2 @ Groupon · Distributed Systems · Cost & Performance
Bengaluru, India
I work on Groupon's Promotions & Dynamic Pricing platform — backend systems handling millions of requests a day across personalized discounting, pricing, and CRM messaging (push / email / banners). Increasingly, a lot of my time also goes into the LLM layer starting to sit on top of all of it.
Most days: Java, Spring Boot, Kafka, Postgres/Bigtable/Cassandra, and GCP/Kubernetes — with detours into Python, Spark/Airflow, MCP servers and AI agents when the problem calls for it. I care about p99 tail-latency, cost efficiency, honest on-call runbooks, and AI that actually moves a business metric.
| Growth |
Checkout upsell system → +14% revenue, +20% order size, +10% conversion Promo-code visibility (Next.js) → ~$500k M1VFM Giftable deal banners under holiday deadlines → ~$600k seasonal impact |
| Performance |
Tail-latency work across IS/MS → errors ↓90%, avg CPU ↓50%, latency peaks ↓75% Audience-load throughput with queue visibility → 2× speed API latency reduction across Promotions surfaces → ~30% |
| Cost |
AWS → GCP migration of core services (zero-downtime) → ~40% perf uplift, ~$500k/yr Rightsizing + Bigtable capacity reduction → ~$120k/yr infra savings Batch heap profiling: 16GB → 8GB with zero regressions |
| AI / LLM |
MCP server for deal search — NL query → structured filters + LLM re-ranking LLM-assisted email subject/body suggestions with offline eval + human-in-the-loop Send-time heuristics with safety guardrails |
| Reliability |
Core QRT responder for pricing, incentives, and CRM incidents Cache freshness, alert revamps, batch retries, and runbooks for peak weeks Bulk Create (CSV → validated campaigns) → 576 campaigns in 23 days |
✍️ Writing — ByteForge
Notes on distributed systems, Java internals, database design, and practical AI engineering. A few pieces I'm proud of:
- 🤖 I Built a Memory System for Claude Code — Here's How It Works
- ⚙️ Mastering Java Concurrency Internals: Atomic Variables, CAS, ABA Problem & LongAdder
- 🗃️ Isolation Levels Made Simple: Balancing Performance and Consistency
- 🏗️ Building Reliable, Scalable and Maintainable Applications
Also writing on: CAP theorem · HTTP versions · pagination patterns · pre-signed URLs · big-data file formats · building a DB from scratch · advanced LLM prompting.
Languages · Java · Python · TypeScript · SQL · C++ Backend · Spring Boot · Play · Akka · Node.js · gRPC · Next.js Data · Postgres · Bigtable · Cassandra · Redis · Kafka · MongoDB · ELK Infra · GCP · AWS · Kubernetes · Docker · Terraform Observability · Grafana · Prometheus · Wavefront
- Checker's King — online multiplayer checkers with WebSockets, ELO, video/chat, spectator mode
- AmbuFast — real-time ambulance tracking with SOS + traffic-control broadcast (Hack36 Top-5)
B.Tech CSE, MNNIT Allahabad (GPA 8.59/10, 2019–2023) · Web System Administrator for the college portal · Summer Intern @ MathWorks · Star Performer @ Groupon
Most of my code lives behind Groupon's firewall — public repos are a small slice. Reach out if you want to talk pricing platforms, performance engineering, or AI in promotions.