Agents Week 2026 | Лекция 2 Memory and Guardrails in LLM-Powered Agents

Speaker: Kirill Mishchenko, Head of the ML Development Group at Yandex Browser In the lecture: We'll discuss the purpose of memory and its different types; We'll discuss RAG and learning as a technique for memorizing information, as well as context management strategies; We'll discuss the importance of Guardrails, the Threat Model, content filters and action limiters, tools that can be used to implement Guardrails, and a little bit about monitoring. Presentation link: https://disk.yandex.ru/d/vWyED4WOXEC07A Intensive course materials that have already been completed: Lecture 1.1: https://youtube.com/live/C1OCgbONSAw?... Lecture 1.2: https://youtube.com/live/VctYHtCap3o?... Learn more about the School of Data Analysis on the website: https://shad.yandex.ru/