Pratik Shrivastava

AI-powered visual fashion search engine that finds visually similar catalog products from an uploaded photo — YOLOv8 for garment detection, DINOv2 for embeddings, and FAISS for similarity search. Generates NLP garment descriptions and retailer links directly from retrieval results, and runs real-time inference on a 20K-image catalog via AWS Lambda.

• FastAPI
• React
• YOLOv8
• DINOv2
• FAISS

Hybrid music recommender blending autoencoder latent similarity, audio features, genre, and popularity into one ranked score, with a per-result panel explaining each signal's contribution. Ships recommendations, mood discovery, genre exploration, playlist building, and a 3D PCA visualization of the embedding space.

• FastAPI
• React
• Vite
• Spotify API

AI-powered book illustration pipeline that extracts text from PDFs via OCR, identifies characters with LLM-based NER and alias merging, and generates page-consistent illustrations by feeding a running visual profile into each prompt. Supports multiple LLM and image-generation backends (OpenAI, Gemini, Cloudflare Workers AI).

• OCR
• LLM NER
• Image Generation
• Supabase

Real-time hand gesture recognition system that detects hand landmarks from live webcam input to classify static poses and dynamic finger movements with low latency.

• MediaPipe
• TensorFlow Lite
• OpenCV

Secure RESTful blogging API with JWT authentication, CRUD for posts, comments, and dynamic tagging on a scalable relational schema.

• Node.js
• Express
• PostgreSQL
• JWT

Full-stack library management system handling book catalog management, borrowing/return tracking, and role-based access for students, librarians, and admins.

• PHP
• MySQL
• Bootstrap