Context Engineering and Memory Patterns for AI Agents
Wednesday, 15 October 2025
18:15 – 20:00
Talent Garden Ostiense, Rome
@CULT hosts “Context Engineering and Memory Patterns for AI Agents” with Danilo Poccia (Chief Evangelist @AWS).
Within the spaces of Talent Garden, @CULT is pleased to welcome a talk dedicated to one of today’s most relevant topics in artificial intelligence: managing context and memory in AI agents.
Danilo Poccia, Chief Evangelist @AWS, will demonstrate - through practical demos - how to avoid context rot and scale reliably, with examples based on Strands Agents and Amazon Bedrock AgentCore.
Why it matters
AI agents represent a new class of applications capable of reasoning, planning, acting, learning, and adapting to achieve user-defined goals with minimal supervision.
Their success hinges on context engineering: deciding which information to include in the model’s context window at each step.
A naïve approach - simply appending the entire conversation history and tool outputs - cannot scale in production, due to:
- the limits of the context window
- degraded performance from context rot
- rising costs as interactions increase
What to expect
Through real-world use cases, the session will present memory management patterns applicable to any LLM or framework, including:
- content summarisation
- memory hierarchies, separating working memory from episodic/semantic stores