BRK310 Breakout Fri 20 Nov 10:15 PST
The evaluation loop for Microsoft IQ
Divya Chirasani, Ahmad Qardahji
Session description
Microsoft's catalogue description
Better context leads to better AI applications, but improvement requires more than trial and error. In this session, we'll build an evaluation loop for Microsoft IQ applications, using evaluations to compare retrieval configurations, context strategies, prompts, and models. Learn how to identify weak spots, measure the impact of changes, and systematically improve application quality. See how evaluations can transform context engineering from guesswork into an iterative, data-driven practice.