AI-powered video interview platform that analyzes uploaded resumes (PDF/DOCX or manual input), detects candidate seniority, and conducts adaptive live interviews. Built with real-time speech-to-text (Whisper), text-to-speech (OpenAI TTS), GPT-based interview logic, LiveKit video/audio streaming, and MongoDB-backed conversation memory for context-aware interview sessions.

A full-stack marketplace to discover and book verified Pandits nearby by location and radius. Built with Next.js and a Node.js/Express API on MongoDB Atlas, featuring JWT auth, geo-based search, puja exploration, multi-step booking, Pandit onboarding, and admin verification. Deployed with Hostinger (frontend) and AWS EC2 + Nginx (backend).

An AI-powered full-stack app that turns GitHub repositories into interactive Mermaid diagrams using Google Gemini. FastAPI backend fetches and analyzes repo files, detects structure and dependencies, then generates flowchart, class, sequence, and module diagrams from file-, function-, or block-level context. Vite frontend with live Mermaid rendering, Tailwind UI, and API proxying for real-time visualization of code architecture.

A full-stack AI analytics chat app that turns natural-language questions into safe, read-only SQL and answers from a live PostgreSQL e-commerce database. Built with a Next.js chat UI and Fastify backend: schema-aware SQL generation via DeepSeek, SQL validation with a read-only DB user, automatic retry on query errors, follow-up conversation context, and responses that include a natural-language answer, result table, auto-selected chart, generated SQL, and execution time—so answers come from real query results, not hardcoded mocks.