Technical Talks

• GenAI Applications

Building agents for real-world knowledge work requires a delicate balance of AI and Human-Computer Interaction (HCI) — one has to understand frontier model capabilities, translate them into a framework for agent behavior (with the right primitives, guardrails, etc), and then place them in an intuitive product surface that is interactive and transparent. The complexity of attaining this balance is magnified for vertical problem spaces that require significant domain expertise to solve for, like law. This talk will share insights and best-practices from building at the bleeding edge of the application layer. We'll explore how to leverage domain expertise to map model problems to legal problems (and importantly, evaluate them), how to create a framework for vertical agents that mirrors human processes, and why, despite LLMs being the star of the show, traditional engineering and machine learning practices are essential for maximizing quality and reliability in production environments.