Software Engineering @ Delhi Technological University (DTU '27)
Building low-level systems, distributed backend infrastructure, and AI-powered engines.
- Algorithmic Problem Solving: Ranked in the top 1.5% globally on LeetCode (Peak Rank #892, Knight 1920) with over 1,450+ problems solved across LeetCode and Codeforces.
- Systems & Internals: Implemented Git's plumbing layer from scratch—rebuilding content-addressable storage, recursive tree binary serialization, and DAG traversals byte-for-byte compatible with official Git objects.
- Distributed & Real-Time: Architected collaborative environments utilizing Yjs CRDTs for conflict-free delta syncs and isolated Docker sandboxes for safe multi-language execution.
- AI & Document Intelligence: Designed production RAG architectures with Pinecone vector databases and Google Gemini, delivering sub-second semantic retrieval with page-level citation grounding.
| Platform | Rating / Title | Global Standing | Solved Count | Profile |
|---|---|---|---|---|
Knight (1920) |
Global Peak Rank #892 (Top 1.5%) | 1,000+ Problems | View Profile ↗ | |
Specialist (1411) |
Div. 2 Peak Rank #1,484 | 450+ Problems | View Profile ↗ |
🌳 GitLight — Git Core Engine & Object Database from Scratch
Byte-for-byte compatible reimplementation of Git's plumbing architecture in Node.js
- Architected a content-addressable storage engine generating valid Git object primitives (
blob,tree,commit) directly consumable by officialgitCLI tooling. - Engineered recursive binary tree encoding to serialize nested directories bottom-up into a persistent Directed Acyclic Graph (DAG).
- Implemented a persistent staging index file and an LCS (Longest Common Subsequence) colored diff engine for line-level change tracking.
- Stack:
Node.js•SHA-1 Hashing•zlib (DEFLATE)•Binary Buffers•CLI Architecture
🔗 GitHub Repository
🤖 CodeSync AI — Real-Time Collaborative IDE with Sandboxed Execution
Multi-user collaborative coding workspace with CRDT-based synchronization and AI code auditing
- Spearheaded collaborative document synchronization utilizing Yjs CRDTs and WebSockets, broadcasting compact binary deltas to eliminate race conditions without full-file rebroadcasts.
- Orchestrated isolated multi-language code execution pipelines inside on-demand Docker containers, enforcing execution timeouts and sandbox isolation.
- Integrated the Gemini API (
@google/genai) to execute automated AST code reviews, delivering contextual optimizations and bug diagnostics based on room language state. - Stack:
React•Node.js•Socket.IO•Yjs (CRDTs)•Docker•Gemini API
🔗 Live Application • GitHub Repository
📄 PDF Genius — Semantic Document RAG Engine & Vector Search
Retrieval-Augmented Generation (RAG) platform with vector embeddings and page-level citation grounding
- Architected an end-to-end RAG pipeline coupling Pinecone vector indexing with Google Gemini embeddings to deliver sub-second semantic retrieval across multi-page documents.
- Engineered document parsing and sliding-window chunking routines using
pdf-parse, enforcing page-level citation references to eliminate hallucinated answers. - Orchestrated a unified Vercel serverless deployment running Express and Vite on same-origin rewrites, with Google OAuth 2.0 authentication and httpOnly JWT session validation.
- Stack:
React•Node.js•Express•Pinecone Vector DB•Google Gemini API•Vercel Serverless
🔗 Live Application • GitHub Repository
| Domain | Technologies |
|---|---|
| Languages | C++ • Python • JavaScript (ES6+) • TypeScript • SQL |
| Backend & Distributed | Node.js • Express.js • REST APIs • WebSockets (Socket.IO) • Yjs (CRDTs) |
| Databases & Vector Search | Pinecone (Vector DB) • PostgreSQL • Redis • MongoDB |
| DevOps & Tooling | Docker • Linux (Bash) • Git • GitHub Actions • Vercel • Postman |

