AI Shopping Assistant

AI Shopping Assistant & Product Discovery Engine Cressi is an AI-powered conversational shopping assistant designed to help users discover products more intuitively. The goal was to build a system that could understand natural language queries, deliver accurate product recommendations, and maintain a fast, scalable backend capable of handling large product catalogs.

I led the development of the end-to-end AI engine, backend services, and product-data ingestion workflows. My responsibility was to ensure the system could retrieve, rank, and recommend products through natural, conversational interactions while keeping the architecture modular and easily scalable.

Implemented a LangChain-based conversational pipeline with OpenAI models to interpret user intent and guide multi-turn shopping conversations.

Integrated Pinecone vector search to handle semantic retrieval of product details and attributes.

Designed a hybrid scoring method combining embeddings, metadata, and keyword filters for more accurate recommendations.

Developed the core backend using Python Flask, structured into microservices for better maintainability.

Built REST APIs for recommendations, browsing, product search, and query handling.

Utilized AWS Lambda + S3 for product media handling, ingestion tasks, and asynchronous processing.

Created a pipeline to aggregate product data from multiple brand websites and APIs.

Normalized and enriched data to maintain consistent structure across thousands of SKUs.

Added validation and ranking logic to ensure top-quality results.

The final system delivered fast, accurate, and context-aware recommendations, enabling users to explore products conversationally. The architecture is scalable, efficient, and optimized for future expansion into broader categories and AI workflows.