Advanced AI Agents with OpenSearch

Building an AI Agent that works in a demo with 3 tools is easy. Building one that works with 100+ tools without blowing up your context window or hallucinating is an engineering nightmare.

The secret to stable agents isn't a smarter LLM, it's better retrieval. In this session, we explore OpenSearch as the deterministic "brain" behind autonomous agents.

We will ditch the hype and focus on the architectural patterns for production agents:

The 100-Tool Problem: You cannot stuff every API definition into your system prompt. We will show how to index tool definitions and use vector search to dynamically retrieve only the top 3-5 relevant tools for a specific user query.

Tool Categorization: Not all tools are equal. We will look at how to handle different tool types. From simple Data Retrieval (RAG) to aggregations and Semantic Actions, we will discuss how to model them effectively in your search index.

Hybrid Memory: Finally, we will apply these same retrieval patterns to Long-Term Memory. We will show why semantic search alone fails for recalling user details, and how to use Hybrid Search (BM25 + k-NN) to build a memory system that actually remembers what matters.

Building advanced RAGs with Elasticsearch \ OpenSearch

RAG (Retrieval Augmented Generation) is built in 3 steps - understanding the query, executing a retrieval step and and generating a response. Each step has it's own unique characteristics and challenges. In this session we'll go through the common pitfalls (and solutions!) for each, and demonstrate how you can easily build an efficient RAG using Elasticsearch or OpenSearch.

WTF Are Embeddings

Embeddings are behind the magic of modern AI - powering search, recommendations, and those eerily accurate chatbots.

But what are they, really?

In this talk, Liza - a regular software engineer, not a data science PhD - breaks it down in plain English using real examples, bad charts, and trusty sciency Python tools.

You'll learn how words, products, or even bananas🍌become vectors in high-dimensional space - and why that matters when building systems like RAG pipelines or semantic search.

Whether you're trying to make your chatbot less clueless or your search results less embarrassing, this crash course will give you the mental model you didn’t know you needed.

Delivered at PyCon IL 2025 and OpenSearch community meetup.
For software engineers getting into AI.