How to Build and Optimize AI Research Agents

In this episode of the Data Exchange podcast, Ben Lorica interviews Jakub Zavrel, CEO of Zeta Alpha, about the distinctions between enterprise search and deep research, particularly within knowledge-intensive companies. Zavrel explains that while enterprise search is foundational, deep research provides more complete answers, saving experts significant time. They discuss the challenges of enterprise search, the importance of customizing deep research tools, and the need for a solid search foundation before implementing deep research. The conversation also covers the use of RAG (Retrieval-Augmented Generation), the integration of structured and unstructured data, and the challenges in customizing and ensuring the quality of deep research. Zavrel also shares insights on using the JEPA (Genetic Pareto Optimization) approach to improve prompt engineering for AI agents, and the episode concludes with a discussion on knowledge graphs and the future of foundation models.