Search Insights & Expertise

Longer queries can challenge lexical search. Learn how important-term boosting can preserve query intent, improve relevance, and better match what users are actually looking for.

Hybrid search doesn’t have to mean merging separate lexical and vector result sets. Explore an alternative architecture that uses semantic search for candidate generation while preserving the relevance, business logic, facets, and control of your existing search platform.

Modern product discovery requires more than traditional search alone. Learn how Large Language Models (LLMs) and BM25-based search engines work together to understand shopper intent, improve search relevance, and deliver more effective ecommerce search experiences.

Building a search proof of concept requires a different mindset than building a production system. This article shares practical OpenSearch tips for rapid prototyping, including vector embeddings, kNN search, hybrid search, indexing strategies, and UI considerations drawn from real-world experimentation.

Traditional search works well when users are looking for specific items. But workflows centered around building cohesive environments, such as set design and scene construction, introduce a different kind of retrieval problem. This article explores how OpenSearch, vector embeddings, and multimodal retrieval techniques were used to experiment with scene-oriented discovery and cohesive item selection.

Search governance adds structure, but it’s only part of the system. As shopper behavior, product catalogs, and business priorities constantly change, search needs to go beyond decisioning and continuously adapt. Governance, control, and optimization work together to keep results aligned with shopper behavior.

This article discusses how Model Context Protocol (MCP) expands AI chatbot capabilities with live data, actionable tools, and smarter workflows to improve customer interactions.

The functionality of the modern blend of AI has some immensely powerful capabilities that offer a great deal of opportunity. LLMs also have some important flaws. Awareness of what AI is, what it’s capable of (and not capable of) is crucial to weighing opportunity cost of IT time. It’s simply too expensive to not be aware of what AI is.

This article explores the challenges and solutions in integrating Java with large language models (LLMs) for search. It details building a proof-of-concept for leveraging popular LLMs like SBERT and OpenAI, using Python-based tools to bridge gaps in Java compatibility. Learn how vector databases, FastAPI, and Chroma enable powerful search functionalities.