I'm a B.Tech Computer Science undergraduate at NIT Hamirpur, focused on backend engineering — building services that have to actually talk to each other correctly, not just return JSON. Most of my project work centers on event-driven architecture, message queues, and distributed systems fundamentals, backed by active DSA practice (500+ LeetCode, rated on Codeforces) and coursework in system design.

I care about understanding why a system is built a certain way — consistency models, failure handling, message delivery guarantees — not just shipping a feature that works on the happy path.

Currently building   →  Multi-service video platform (Spring Boot + RabbitMQ + FFmpeg)
Currently learning    →  Distributed consensus, system design at scale
Actively practicing   →  DSA across LeetCode (500+) and Codeforces
Open to               →  Backend engineering internships & SDE roles

Languages

Backend & Messaging

Data & Infra

Tools

A multi-service video upload and transcoding system. The core API (streamingv2) issues presigned S3 upload URLs, persists video metadata, and publishes a processing job to RabbitMQ. A transcoding-worker service consumes that job, runs multi-bitrate FFmpeg transcoding, and uploads the HLS output back to S3; a third notification-worker service independently consumes the "processed" event and emails the uploader — three deployable Spring Boot services that only share a message contract, plus a Next.js frontend for upload/playback.

Spring Boot RabbitMQ PostgreSQL AWS S3 + CloudFront FFmpeg Spring Security Next.js

→ Producer/consumer chain across 3 services, structured AMQP message DTOs, presigned-URL upload flow, adaptive-bitrate HLS, video status state machine.

Two standalone REST APIs built to get authentication and authorization right: JWT-based auth with OTP email verification and rate limiting (auth/), and a role-based file upload service with Multer, PostgreSQL via Drizzle ORM, and per-user/admin ownership checks (file_upload_system/).

Node.js Express MongoDB PostgreSQL Drizzle ORM JWT Zod

→ Layered architecture (controller / service / middleware / validator), RBAC middleware, request validation, rate limiting.

A backend-driven mock technical interview system built around five chained LLM agents — problem understanding, approach evaluation, coding, code review, and final report generation — each with an externalized prompt file, orchestrated as a sequential pipeline where a deterministic controller judges scores and the agents only evaluate. Ships with a Monaco-powered code editor on the frontend. Built on Gemini.

Node.js Express MongoDB Gemini API React Monaco Editor

A dual-pipeline backend serving citizen-facing fraud Q&A and investigator-facing fraud network analysis, with a Neo4j graph layer for entity/relationship modeling and Pinecone hybrid-search retrieval. Permission enforcement is backend-side by design — an LLM's classification is never treated as an authorization mechanism. Early-stage: retrieval is the fully implemented slice, the graph/risk layers are still scaffolded.

Python Neo4j Pinecone Gemini API