HarshitRaj
Backend engineer focused on reliability, observability, and calm execution.
Harshit Raj
I'm a systems-oriented backend engineer. I work on high-scale pipelines, distributed services, and the operational tooling that keeps them predictable in production.
I've built and operated Kafka-based systems, microservices, and observability stacks at AdPushup (AdTech) and Amazon Transport Tech. I care about correctness under load, good alerting hygiene, and reducing incident noise.
I'm intentionally optimizing for sustainable pace and calm execution. I prefer clear scope, async-friendly workflows, and boring reliability over heroics.
Work History
Systems work: reliability, correctness, and operations. Emphasis on outcomes and production maturity.
- Owned backend features for order exchange, returns, and OTP delivery; handled production issues (service connectivity, queue delays, timeouts).
- Containerized a 250MB+ Lambda service into a Docker image; improved deployment consistency and reduced cold-start latency by ~40%.
- Reduced database footprint 24GB → 600MB by purging millions of stale order events via compliant 30‑day TTL; cut storage cost and eliminated false-positive signals.
- Migrated a large Lambda service to AWS CDK and enabled integration tests in Gamma to strengthen pre-prod validation.
- Replaced unbounded logs with a 3‑month CloudWatch retention policy using IaC to codify governance and reduce storage growth.
- Engineered a Kafka-based feedback infrastructure processing 3.2B+ daily requests with high reliability and fault tolerance.
- Built a multi-threaded ingestion pipeline for Google Ad Manager; reduced report processing lag by ~90%.
- Implemented a Kafka Streams anomaly detection service to identify traffic deviations in real time.
- Built a Prometheus + Grafana + PagerDuty observability stack; improved MTTR by ~50%.
- Shipped ad-delivery optimizations improving auction efficiency at scale; contributed to ~20% average eCPM lift across key ad units.
Contributions & Maintenance
Small, targeted changes that improve correctness, UX, and long-term maintainability.
Fractional-offset scheduling, UTC parsing refactor, and docs improvements.
Added unit tests and CI coverage for spinner components (Go).
UX fixes and stability improvements for the v1.0 release (Go + HTMX).
A bit personal
A few things outside work. If you want the longer version, it's on the Beyond Code page.
Consistency and recovery. Clear head, better work.
New places, long walks, fewer screens.
Systems notes, reliability, and practical engineering lessons.
Side projects for learning and utility — not hustle.