Roca-III

Generative AI prototypes are easy to build—but making them enterprise-ready is not. In this session, we will explore how LangFlow, a visual orchestration framework built on LangChain, can be integrated with AWS-native services to design scalable, secure, and production-grade GenAI workflows.

We will cover how to deploy LangFlow on AWS (ECS/Fargate or EKS), integrate Amazon Bedrock as the LLM provider, use Amazon S3 for document ingestion, and apply IAM, networking, and secrets management best practices. The session will also discuss architectural patterns for extensibility, observability, and governance—key concerns for real-world enterprise adoption.

Attendees will leave with a clear understanding of where LangFlow fits in the AWS GenAI ecosystem and how to move beyond PoCs toward production.

I’ll introduce AWS Nuke for systematic resource cleanup, a hard reset button for AWS! We’ll dive into exactly how to build a production-ready solution by elevating the standalone CLI tool into an automated service using a custom AWS Architecture.

You’ll learn how to enable your team, see the automation in action via a demo, and discover why you absolutely shouldn’t eat the brown M&M’s.

How do we know that our terraform in Git is in sync with Deployments?
How do we ensure automatic Terraform deployments?

As Infrastructure as Code (IaC) adoption accelerates, maintaining synchronization between version-controlled Terraform configurations and deployed infrastructure has become increasingly critical.

This presentation explores the challenges of ensuring GitOps alignment and examines the most prominent solutions for automating Terraform workflows.

Whether you’re managing a startup’s infrastructure or orchestrating enterprise-scale deployments, this session will equip you with the knowledge to make informed decisions about automating your Terraform workflows while maintaining the critical link between code and deployed resources.

Food and drinks are two of my major hobbies, and I love mixing cocktails at home. But when I host friends, I always end up as the menu. People keep asking what I have and what I can make.

In this talk I will show how I built an AI bartender on AWS that lets guests interact with a drink menu using natural language. Instead of listing options or explaining what’s available, guests can ask for things like “something fruity” or “a gin drink”, and the AI Bartender respond with suggestions powered by Amazon Bedrock.

We will go deep into how I built the AI agent with Amazon Bedrock using Strands Agents, and AgentCore providing conversation memory, identity, and a gateway for tool access. I’ll show how streaming responses, with Amazon API Gateway, are implemented for a great chat experience. MCP tools play a important role providing the agent with data, like ingredients, and menu stored in Aurora DSQL. Finally how serverless and event-driven patterns are used to decouple AI processing from APIs.

This is a technical, demo-heavy session with code walkthroughs and architecture deep-dives. We will explore agentic AI with tool calling workflows, conversation memory, and event-driven AI integration. You’ll walk away with insights on how to build intelligent, context-aware AI assistants that work with real data on AWS.

In 2025, Agentic AI has demonstrated its ability to address a wide range of business domains — from financial services to marketing and retail. As successful pilots and early deployments deliver tangible value, organizations are increasingly pushing agentic systems from experimentation into production environments.

While Agentic AI promises autonomy and advanced reasoning, it also introduces a new class of production risks that traditional software systems — and even earlier Generative AI applications — do not face. These risks often surface only after early success, when autonomous behavior, non-determinism, and system-level interactions begin to scale.

In this talk, we examine the most common failure modes observed in real-world agentic AI deployments and discuss how to address them through robustness-aware system design. We explore architectural patterns, guardrails, and observability layers that help teams move beyond fragile prototypes and build agentic systems that are reliable, governable, and production-ready on AWS.