AI-Powered Virtual Try-On and Stylist
FastAPI, Python, React, Gemini API. Architected a modular FastAPI backend handling 500+ concurrent requests with under 2-second response time, implementing asynchronous processing, structured logging, and event-driven status callbacks. Reduced failed AI requests by 40% through robust input validation, structured error-handling pipelines, and systematic debugging of language model prompt orchestration modules. Built end-to-end image pre-processing and response transformation scripts, demonstrating clean code practices, OOP design patterns, and comprehensive API documentation.
F1 Telemetry and Strategy AI Dashboard
React, TypeScript, Scikit-learn, Gemini API. Designed a real-time data ingestion pipeline processing 10,000+ telemetry data points per session, applying data cleaning, feature engineering, and transformation for downstream ML modules. Implemented K-Means clustering to classify 3 driving styles with 85% accuracy and integrated results into Gemini API for automated natural-language race strategy report generation. Managed the end-to-end project life cycle using Git version control and Agile task tracking, ensuring reproducible builds and consistent software delivery.
AI-Powered Skill Bartering Platform
MERN Stack, Python, WebRTC, PostgreSQL, Redis. Architected a full-stack platform serving 200+ active users with microservices architecture; built a Python-based AI matching engine using Gemini API, achieving 90% user satisfaction. Deployed Redis-backed caching to scale to 1,000+ concurrent connections, optimizing throughput and reducing API latency for real-time chat and video sessions. Integrated WebRTC via REST API signaling, reducing server bandwidth by 70% while maintaining under 150 ms latency, demonstrating multi-service API integration and system design skills.