In this tutorial, we’ll walk you through building an advanced product search engine using LanceDB. We’ll start with a raw dataset of fashion products and use LanceDB’s powerful features to engineer new features, generate embeddings, and build a sophisticated search and retrieval pipeline.
The tutorial is divided into two parts:
Part 1: Feature Engineering with LanceDB and Geneva: In this part, we’ll focus on the crucial process of feature engineering. We’ll use LanceDB and its Geneva feature engineering framework to enrich our data with meaningful features that will power our search engine.
Part 2: Inference and Retrieval with LanceDB: In this part, we’ll build the inference and retrieval pipeline that uses the features we engineered in Part 1 to provide a powerful and intuitive search experience. We’ll cover query routing, hybrid search, and reranking to build a state-of-the-art search engine.