Improving LLM prompts using data-driven feedback optimization - Mikhail Sveshnikov

​Outline:

• ​Overview of prompt optimization challenges and common approaches (manual iteration, few-shot learning, etc.)
• ​How Evidently AI's prompt optimization works: using feedback from mistakes on real data to iteratively improve prompts
• ​Live demonstration: optimizing a classification prompt step-by-step, showing how errors are identified and used to refine the prompt
• ​Q&A and discussion: best practices, when to use different strategies, and practical consideration

​ ​​About the Speaker:

​Mikhail Sveshnikov is an AI engineer at Evidently AI with 10+ years in ML and MLOps, focused on building developer tools for reliable and measurable AI in production. ​

Join our Slack: https://datatalks.club/slack.html