Solutions Backend Engineer at Razorpay — working in payments infrastructure where correctness isn't optional and failures have real financial consequences.
Day to day: debugging live transaction failures, stabilizing merchant integrations across Payout and Fund Account Validation APIs, and contributing to platform reliability at scale.
Before Razorpay, I built a multi-tenant workflow automation platform from scratch at Brane Enterprises — microservices, Spring Boot, REST APIs, query optimization across the full lifecycle.
I gravitate toward problems where async systems, failure recovery, and distributed consistency actually matter.
Java · Spring Boot · MySQL · Redis · Resilience4j · JWT · HMAC · Docker · AWS
A fault-tolerant, distributed-ready backend system that validates bank accounts and VPAs through external payment provider APIs — built to mirror how real fintech backends handle async flows, distributed consistency, and failure recovery.
I recently upgraded this significantly and would genuinely love for you to take a look 👇
What's under the hood:
- Fully asynchronous validation pipeline — API returns immediately, workers process in the background
- Multi-instance safe: designed to run across 5+ Docker containers with no race conditions on state transitions or retries
- Distributed rate limiting via Redis + Bucket4j — limits enforced uniformly across all instances
- Circuit breaker (Resilience4j) protecting the system from cascading failures when providers degrade
- Idempotency controls — safe retries with no duplicate processing, critical for financial correctness
- State-driven lifecycle (
INITIATED → PROCESSING → COMPLETED/FAILED) preventing inconsistent updates - Webhook ingestion secured with HMAC signature verification
- Scheduler-based reconciliation that auto-recovers validations when webhooks are delayed or never arrive
- Fully containerized with Docker Compose, deployed on AWS EC2
This project directly mirrors the Fund Account Validation system I work with at Razorpay — built to understand it from the inside out.
📂 View Repository · 🎥 Watch Demo
Python · FastAPI · LangChain · ChromaDB · Gemini · Azure OpenAI (GPT-4o)
A multi-tenant, enterprise-grade Retrieval-Augmented Generation backend — built with a modular monolith architecture and production-level reliability patterns.
What's interesting about this one:
- RBAC enforced at the vector database layer via metadata filtering — external users physically cannot retrieve internal knowledge chunks, regardless of what they query
- Circuit-breaker model routing: if Gemini 2.5 Pro hits rate limits, traffic auto-falls to Gemini Flash, then Azure OpenAI GPT-4o — zero downtime
- Self-healing knowledge injection: admins can push hardcoded Q&A corrections directly into the vector store without any model retraining
- Clean client-server separation — Streamlit UI is a pure presentation layer, zero AI or DB logic on the frontend
- How distributed systems stay consistent under failure — not just in theory, but in the specific ways payment systems break in production
- The gap between "it works" and "it's observable, recoverable, and safe to operate" — and how to close it through better architecture choices
- How large-scale platforms like Razorpay's Payouts handle correctness guarantees at volume
BU Open House — Rookie MVP · Razorpay · 2025
Awarded for outstanding backend contribution and measurable impact within the first 3 months at Razorpay.
