Modern software systems operate in complex, dynamic environments where failures are inevitable. Traditional monitoring and manual incident response are no longer sufficient to ensure resilience or customer satisfaction. This talk explores how to design and implement self-healing software systems by combining telemetry data with an AI-driven agentic approach. We’ll start by examining how high-quality telemetry forms the foundation for detecting anomalies and predicting failures. Next, we’ll show how modern GenAI (LLMs) can transform this telemetry into actionable insights for AI agents that interpret data, pinpoint root causes, and apply automated fixes. Through a practical, real-world example, you’ll see how telemetry and AI work together to create adaptive feedback loops that continuously improve system reliability, while freeing engineers from repetitive operational tasks.

Recreating the unique state of a cloud app environment is notoriously complex, making it difficult to diagnose bugs, collaborate with teammates, or even just pick up where you left off. Join us as we deep dive into LocalStack’s powerful state management and persistence features, exploring how to maintain, snapshot, and seamlessly share your local cloud environments using core features like persistence, Cloud Pods, and state files. Learn how to restore a previous state, share preconfigured environments with colleagues for collaboration or onboarding, or restore preset services and data for functional tests in CI. We'll also showcase LocalStack's new Model Context Protocol (MCP) integration, revealing how AI can manage, inspect, and automate your local cloud state.

An agent will write you a CDK stack, a Terraform module, or a stack of IAM policies in seconds.
Whether any of it works is a separate question, and the usual way to find out is to deploy to a real AWS account and watch what breaks.
In an agentic workflow, that means giving AI access to a public cloud account, racking up costs on the AWS bill, and waiting for provisioning to complete every time you push new code to the environment.

The rise of agentic AI in the software delivery lifecycle creates a dilemma with high-stakes implications.
As agents create new applications at an unprecedented rate, how do you integrate security without slowing down delivery?