Backend-Focused Full Stack Engineer
Building systems
that scale.
3+ years building full stack products — with a deep focus on distributed backends, microservices, real-time engines, event-driven architecture, and AI-powered systems including RAG pipelines, agent frameworks, and real-time voice agents. Based in Bengaluru, India.
I'm a full stack engineer with a strong backend focus — specialising in Node.js, TypeScript, and distributed systems. Over 3 years I've built and shipped production products across the entire stack, though my deepest work lives in the backend: re-architecting monolithic platforms into microservices, designing real-time tracking engines, and building event-driven workflows that hold up under load.
On the frontend I'm comfortable in React — I've built and delivered full features using Redux, Tailwind, and shadcn/ui, and have stepped in as the frontend owner when needed in lean teams. I care about shipping complete, polished products, not just APIs.
At Cloudesign, I independently owned backend development for CloudTrack — a multi-tenant logistics SaaS — covering everything from trip management and real-time tracking to a live auction/bidding engine and billing workflows. Outside of work, I build projects like RideFlow: a fully distributed ride-hailing backend modelled on real-world scale.
More recently, I've been building AI systems with Python, FastAPI, LangGraph, RAG, embeddings, vector databases, and real-time streaming. My current AI work includes VoicePilot AI, an AI voice receptionist with a stateful agent workflow and retrieval-backed company knowledge.
I write on backend engineering and system design on Medium. Always up for conversations about architecture, system design, distributed systems, or practical AI engineering.
- Re-architected a legacy monolithic logistics platform (CloudTrack) into a scalable, multi-tenant B2B SaaS product using NestJS, TypeScript, Redis, and MongoDB — implementing tenant isolation, RBAC, and feature flags.
- Designed and maintained production REST APIs with request validation, consistent error handling, API versioning, and JWT-based authorization.
- Built the core Trip Management system — creation, driver assignment, OTP-based pickup verification, and full status lifecycle transitions.
- Implemented dual-mode real-time tracking via WebSocket (driver app) and SIM-based GPS (Twilio) for live truck location on the frontend.
- Built a fault-tolerant real-time auction/bidding engine with WebSockets, async event processing, and retry/consistency controls to prevent race conditions under concurrent bids; extended it to a WhatsApp-native workflow for non-web transporters.
- Optimized high-throughput APIs through MongoDB query/indexing tuning and Redis caching on high-frequency read paths, reducing API response times by 40–60% in production.
- Built a high-volume Notification Service spanning Email, WhatsApp, SMS, and Push using asynchronous messaging, retry mechanisms, and event-driven workflows for reliable delivery.
- Built end-to-end billing, invoicing, and agreement/contract modules with customer-specific pricing rules, contract validation, and financial workflow automation across tenants.
- Delivered customer-facing features and real-time dashboards that improved operational visibility across logistics workflows.
- Led code reviews, mentored junior engineers, conducted 45+ technical interviews, and established engineering best practices for the team.
- Delivered Denim Store — a full-stack e-commerce platform with Razorpay payments and an admin dashboard for product, order, and coupon management.
- Delivered Event Crest — a vendor booking platform with real-time chat (Socket.IO), Stripe payments, and live streaming via ZegoCloud.
Multi-tenant B2B logistics SaaS platform built at Cloudesign Technology Service. Re-architected from a legacy monolith into a scalable microservices backend. Owns the full logistics operational lifecycle — trip creation and management, OTP-based pickup verification, dual-mode real-time truck tracking (WebSocket + Twilio SIM GPS), a live auction/bidding engine for trip assignments, a WhatsApp-native bid workflow for non-web transporters, and end-to-end billing and agreement modules. Redis caching reduced API latency by 40–60% in production.
Production-grade distributed backend for a ride-hailing platform. 7 independent NestJS microservices — API Gateway, IAM, Driver, Geo, Matching, Trip, and Payment — each with its own database and event-driven inter-service communication via RabbitMQ. Features a driver matching engine with sequential/parallel offer modes, full trip lifecycle state machine, billing saga with Dead-Letter Queue fallback, MongoDB geospatial indexing, surge pricing, and replay-resistant request design.
Vendor booking platform connecting users with event organizers. Real-time chat via Socket.IO, Stripe payment integration, and live streaming capability via ZegoCloud. Built with React.js, TypeScript, and Node.js.
Full-stack e-commerce platform for men's and women's denim fashion. Secure checkout via Razorpay, admin dashboard for product, order, and coupon management. Built with Node.js, Express.js, MongoDB, and EJS.
Production-oriented AI voice receptionist backend for real-time phone conversations. Built a streaming Speech-to-Text → LLM → Text-to-Speech pipeline with FastAPI, WebSockets, LangGraph, Redis, and Deepgram. The agent uses a stateful graph with memory, routing, tool execution, and answer nodes, with ChromaDB-powered retrieval for company-specific knowledge. Designed for low-latency conversational flows and extensibility toward Twilio, ElevenLabs, PostgreSQL, and production observability.
Retrieval-Augmented Generation layer for VoicePilot AI. Company knowledge is converted into embeddings, stored in ChromaDB, retrieved through similarity search, and injected into the LLM prompt alongside conversation history. Implemented the retrieval flow so the assistant can answer using relevant business knowledge instead of relying only on model memory.
Stateful agent architecture built with LangGraph. The workflow separates memory, routing, tool execution, and response generation into explicit graph nodes, making the system easier to extend with business tools, retrieval, conversation memory, and external services while keeping the execution flow observable and testable.
Production-style AI agent framework built with LangGraph, FastAPI, ChromaDB, and OpenAI/Gemini. Implements session-based memory, tool calling, and conditional routing, with multi-step agent workflows designed using StateGraph for dynamic tool selection and context-aware responses. Integrates vector search via ChromaDB for semantic document retrieval and Retrieval-Augmented Generation (RAG) question answering.
Distributed ride-matching subsystem designed around real-world ride-hailing workflows. Includes geospatial driver discovery, automatic search-radius expansion, sequential and parallel driver offers, live trip state transitions, and idempotent request handling to prevent duplicate operations such as double charges. Structured logging and distributed request tracing across services simplify debugging and production issue diagnosis.