AI-powered products, GenAI applications, and enterprise systems built across AI, fintech, automotive, and travel domains.
AI Project
AIFull-stack RAG (Retrieval-Augmented Generation) application that lets users upload documents and chat with their data using LLMs. Features vector embeddings, semantic search, and conversational AI with citation-backed responses.
End-to-end GenAI product from embedding to conversational UI
AI Project
AIAutonomous agent system that breaks complex tasks into sub-tasks, delegates to specialized AI agents, and synthesizes results. Built with tool-use patterns, MCP integration, and human-in-the-loop controls.
Multi-agent orchestration with autonomous task decomposition
Expedia Group
Enterprise visual workflow editor with drag-and-drop interface. Led migration from Vue.js to React, establishing new standards for component architecture across the platform. Foundation for AI-driven workflow automation.
3x developer velocity improvement
Open Source
Interactive Windows 95-themed portfolio experience built with Astro 5, React 19, and Three.js. Features a full desktop environment with draggable windows, a terminal, 3D skill galaxy, easter eggs, and CRT effects.
Full interactive desktop experience in the browser
Tekion Corp
Multi-language F&I module serving thousands of automotive dealerships. Built with full WCAG compliance and internationalization support for global market expansion.
Serving thousands of dealerships
BYJU'S
High-reliability payment validation system processing over $10M in monthly transactions with 99.9% uptime. Built with robust error handling and real-time monitoring.
$10M+ monthly transactions processed
Open Source
Performance-optimized portfolio built with SvelteKit and Svelte 5. Features AI-ready content with llms.txt, structured data for AI search engines, RSS feeds, and a technical blog focused on AI/GenAI topics.
98+ Lighthouse score, AI-indexed with llms.txt
Personal Project
CLI tool that analyzes React component render performance and identifies unnecessary re-renders, large bundle contributions, and missing memoization opportunities.
Identified 40% redundant renders in test projects