Krish Kumar
Featured project
MemoryVerse AI '26
MemoryVerse MemoryVerse is an AI-powered document identity and knowledge platform. Users upload their academic & professional documents (PDF/DOCX/TXT), and the platform automatically extracts metadata, visualizes skill connections, and provides RAG AI chat over their documents. š ļø Tech Stack & Key Choices Backend: Python 3.11 + FastAPI + SQLite (SQLAlchemy) ā Lightweight REST API. AI & Search: Google Gemini API (gemini-flash-latest for generation/RAG, models/gemini-embedding-001 + pure-Python cosine similarity for vector search). Frontend: Vanilla HTML5, CSS3, JavaScript & Canvas API ā Zero heavy frameworks (React/Vue), keeping the bundle under 20MB. Security & Auth: Users: JWT Bearer Tokens + bcrypt password hashing. Admin Docs: /api/docs protected via HTTP Basic Authentication. Deployment: Vercel Serverless via GitHub CI/CD. š Main Features AI Document Ingestion: Auto-extracts titles, summaries, dates, organizations, and skills. Interactive Knowledge Graph: Canvas-based node graph showing relationships between documents. Conversational RAG Chat: Natural language QA strictly grounded in uploaded documents. Career Timeline: Auto-generated chronological career history. Theme Support: Premium Dark / Light mode toggle.