Available for opportunities.
been here for years
A passionate and dedicated software engineer with a strong foundation in computer science and a knack for problem-solving. Experienced in building web applications using modern technologies like React, Next.js, and Node.js. Eager to learn and adapt to new challenges in the tech world. Available for opportunities.
Stealth AI Startup
TechXtract
Amazon ML Summer School
Social (Formerly Script Foundation)
JavaScript
Gemini AIChrome Extensions APIGitHub
Project ManagementCode ReviewTeam CollaborationMentorshipIBM-CSRBOX
Maharaja Agrasen Institute of Technology
School of Excellence, Rohini
School of Excellence, Rohini

Winnow
Open-source RAG prompt compression middleware. Sits between your vector DB and LLM, compressing retrieved context using LLMLingua-2 token-level scoring guided by your query - cutting token costs by ~50% with less than 3% accuracy loss. OpenAI-compatible proxy, LangChain integration, self-hostable via Docker, pip installable.

Credit Card Fraud Detection
Machine learning project comparing six classification algorithms to detect fraudulent credit card transactions on highly imbalanced dataset. XGBoost achieved 97.71% AUC while LightGBM with 5-fold CV provides robust production model at 95.84% AUC. Analyzed 284,807 transactions with comprehensive EDA and feature importance analysis.

Winnow
Open-source RAG prompt compression middleware. Sits between your vector DB and LLM, compressing retrieved context using LLMLingua-2 token-level scoring guided by your query - cutting token costs by ~50% with less than 3% accuracy loss. OpenAI-compatible proxy, LangChain integration, self-hostable via Docker, pip installable.

Chat-ly
Full-featured real-time messaging platform with instant chat powered by Socket.IO, WhatsApp-style status stories, AI-powered smart reply suggestions via Google Gemini, online presence tracking, and in-app notifications. Built as a Turborepo monorepo with Next.js 16, Express, and PostgreSQL.

Credit Card Fraud Detection
Machine learning project comparing six classification algorithms to detect fraudulent credit card transactions on highly imbalanced dataset. XGBoost achieved 97.71% AUC while LightGBM with 5-fold CV provides robust production model at 95.84% AUC. Analyzed 284,807 transactions with comprehensive EDA and feature importance analysis.

NeoBazaar
E-commerce website built with Next.js featuring Sanity CMS for content management, NextAuth for authentication, and Stripe payment gateway integration. Deployed on Vercel with responsive design and modern UI components.
Selected for the Big Data course under Samsung Innovation Campus (SIC).
Ranked 3rd in a Quira quest and placed Top 10 in 2 additional quests - competing in open-source coding challenges on quira.sh.
5+ hackathons competed - build fast, break things, ship relentlessly. Top 20 @ SuperMind 3.0, SIH qualifier.
Team's startup idea selected under NASSCOM's Tech for Change initiative.
Project admin for ThinkDSA under SSOC-4, managing 10+ contributors and shipping features in the open.
Selected for Amazon ML Summer School - deep learning, reinforcement learning, and production ML pipelines.
Want to collaborate? Have an opportunity? Let's talk.
[email protected]