PROJECTS
TalentLens — LLM-Powered Candidate ATS
A generative-AI-driven ATS that intelligently evaluates resumes line-by-line, rewrites vague statements, and predicts best-fit roles and companies. Built with a hybrid retrieval pipeline to ensure context-aware and truthful rewriting.
- Built an AI-powered ATS that critiques resumes line-by-line, rewrites vague points, and suggests best-fit roles and companies using RAG + Llama 3.1.
- Implemented hybrid retrieval (BM25 + embeddings) to ground feedback in real resume and JD context, ensuring truthful and specific rewrites.
- Designed structured LLM pipelines in LangChain for JSON-based outputs, enabling consistent reasoning and line-level citations.
- Generated JD-tailored resume redrafts with factual placeholders and predicted top-3 matching roles/companies using semantic similarity.
ChatApp
A real-time chat application built with JavaScript and Socket.IO that enables instant messaging with room-based communication.Features user management, dynamic room creation, and bot notifications.
- Real-Time Messaging Engine, Instant message delivery using Socket.IO event-driven architecture
- Message formatting with timestamps and sender metadata
- Dynamic room creation/joining system
- Broadcast messages to specific rooms
- Live user lists per room
- Automatic room updates on user changes
Unsplash Photo Gallery
A modern React application that leverages the Unsplash API to provide an elegant image search experience. Features real-time search, infinite scrolling, and responsive gallery layouts - demonstrating clean React architecture and API integration patterns.
- Instant search-as-you-type functionality
- Debounced API requests for performance optimization
- Dynamic loading of images on scroll
- Pagination management with Unsplash API, Loading indicators during data fetch
- Masonry-style image grid using CSS Grid/Flexbox
- Mobile-optimized responsive design
- Hover effects with image metadata display
- Click-to-expand image details, Display photographer information and EXIF data