In open commerce ecosystems like ONDC, users often struggle to find relevant products across thousands of sellers using traditional keyword-based search. Typing vague or conversational queries like “white kurta for women under 1500” or “bag for travel and laptop” usually returns poor or inconsistent results. There is also no support for visual references or voice input, making product discovery less intuitive and more frustrating. This project solves that by offering a smart, AI-powered discovery engine that leverages both NLP and computer vision models to understand user intent and context. It uses the MiniLM (all-MiniLM-L6-v2) transformer model for semantic text understanding, FAISS for fast similarity search, CLIP (ViT-B-32) for image-based product matching, and Whisper for transcribing voice queries into text. These models allow users to search by typing, speaking, or uploading an image, and receive highly relevant fashion product suggestions. The system also auto-detects smart filters like gender, usage type, and price, and curates the results into user-friendly categories like “Top Picks” or “Budget Friendly”. This makes product discovery significantly more intuitive, efficient, and personalized for users while improving product visibility and engagement for sellers.