🚀 Mrigank Singh | AI-Integrated Portfolio

👋 Overview

Welcome to my personal portfolio! This isn't just a static showcase of my work—it's a Full-Stack AI Application.

Beyond displaying my projects (DASES, MealMatch) and my 3 patents, this site features a custom RAG (Retrieval-Augmented Generation) Chatbot. "Mrigank AI" can answer questions about my technical skills, experience, and background in real-time, powered by a vector database and Google's Gemini LLM.

✨ Key Features

• 🤖 AI Assistant (RAG Chatbot):
  • A persistent, context-aware chatbot trained on my resume and technical documentation.
  • Tech: Google Gemini 2.5 Flash Lite + Pinecone Vector DB.
  • UX: Features "typing effect" streaming, Markdown rendering, and session persistence.
  • Architecture: Desktop Sidebar / Mobile Bottom Drawer.
• 🎨 Modern UI/UX:
  • Glassmorphism design language using Tailwind CSS.
  • Smooth animations powered by Framer Motion.
  • Fully responsive layout for all devices.
• 🛠️ Real-World Engineering:
  • Cold Start Handling: Implemented "Health Pings" to wake up the serverless backend immediately on site load.
  • Security: CORS configured backend and clean environment variable management.

🛠️ Tech Stack

Frontend (This Repo)

• Framework: React (Vite)
• Styling: Tailwind CSS, PostCSS
• Animations: Framer Motion
• Icons: Lucide React
• Markdown: react-markdown with remark-gfm
• Deployment: Vercel

Backend (Separate Repo)

• API: FastAPI (Python)
• LLM: Google Gemini 2.0 Flash Lite Preview
• Vector DB: Pinecone (768 Dimensions)
• SDK: Google GenAI SDK v1.0
• Deployment: Hugging Face Spaces

🧠 AI Architecture

The "Mrigank AI" chatbot works on a Retrieval-Augmented Generation (RAG) pipeline:

1. Ingestion: My resume and project docs are chunked and embedded using gemini-embedding-001.
2. Storage: Vectors are stored in a Pinecone index.
3. Query: When you ask a question, the backend searches Pinecone for the most relevant context.
4. Generation: The context + your question are sent to Gemini 2.5 Flash, which generates a grounded response.
5. Response: The answer is sent back to the frontend and rendered with a typing animation.

📬 Contact

• Mrigank Singh
• Email: [mriganksingh005@gmail.com]
• LinkedIn: [https://www.linkedin.com/in/mrigank005]
• GitHub: [https://github.com/Mrigank005]

---

© 2026 Mrigank Singh. Built with ❤️ and a lot of caffeine.