A Builder’s Guide to Agentic Search & Retrieval with Doug Turnbull & John Berryman
Imagine a system where users can express complex, multi-step needs in plain language, and an intelligent agent uses a suite of tools to not only find information but to synthesize, analyze, and even act on it.
This is the new frontier, and it requires a fundamental shift in how we think about building, evaluating, and designing search, retrieval and product experiences.
Join host Hugo Bowne-Anderson for a deep-dive conversation with search veterans Doug Turnbull (Led Search at Reddit + Shopify; Wrote Relevant Search + AI Powered Search) and John Berryman (Early Engineer on Github Copilot; Author of Prompt Engineering for LLMs). With decades of combined experience building search systems for the enterprise, they are on the front lines of this evolution.
In this live session, we'll provide practical playbook for building the next generation of search. Come with questions!
In this session, you will learn:
An Incremental Path to Agency: Discover a practical, three-stage roadmap to evolve your existing search system into a powerful agentic experience without a massive, risky overhaul.
RAG is Not Just Vector Search: Uncover why relying solely on embeddings is a path to an "80% solution" and learn how to master hybrid systems that blend the power of classic lexical search with modern semantic understanding.
Evaluation in a Dynamic World: Move beyond static "golden datasets." Learn battle-tested strategies for evaluating your search systems, including the promises and critical pitfalls of using "LLM as judge."
The Future of UX: Explore key design principles for conversational search, ensuring your users have an intuitive and powerful experience as the search bar morphs into a chat window and beyond.
A Glimpse of the Future: Join a forward-looking discussion on how agents will reshape the internet, turning websites into data sources and creating a personalized, on-the-fly web for every user.
Register to join live or get the recording afterwards.