We built an AI-powered search agent that turns messy, natural language shopping requests into precise filters like category, price, color so customers actually find what they want. Using a large language model with retrieval-augmented generation and catalogue-aware embeddings, our system understands user intent in real time and grounds it in actual product data.
In tests with 50,000 fashion queries, we cut “no results” by 47% and improved filter accuracy by over 20%, paving the way for truly conversational, personalized e-commerce search.