Tech Stage

Sessions

WELCOME01

Room: Tech Stage | Time: 14:00

ACD401

Room: Tech Stage | Time: 14:10

On a platform with 8+ million accounts built on top of a Cognito User Pool we ran into several annoyances with the service. Since we were not using it in the traditional way, we were able to completely replace it with a custom solution using DynamoDB. I will cover the reasons why we wanted to move away from Cognito, which challenges we faced and how we solved them with zero downtime or user impact.

Robert Slootjes

Director of Platform Technology @ Superstruct Digital / ID&T

ACD402

Architecting Secure and Observable AI Agents Using Amazon Bedrock AgentCore

Room: Tech Stage | Time: 14:30

AI agents are moving from experimentation to production, making identity and observability critical foundations for secure agentic AI. This session explores how Amazon Bedrock AgentCore can be used to implement authentication, authorization, tracing, monitoring, and governance for AI agents. Learn practical architectural patterns for building secure, transparent, auditable, and production-ready AI agents that can operate confidently at scale.

Dmytro Lesnyak

Head of AWS & GCP at Capgemini Finland

ACD403

From LLM Bypass to Stolen Secrets: Tracing a Multi-Layer Attack in AWS

Room: Tech Stage | Time: 14:50

A prompt injection can end with a stolen secret and a cloud breach. This talk traces one multi-layer attack across five signal domains in your AWS environment: L7/API traffic inspection, workload runtime, network flows, EKS audit logs, and CloudTrail. No single source tells the full story. We will look at where the industry, AWS included, is heading: correlating runtime signals with AI-assisted investigation to turn noise into a real assessment.

Bruno Gabriel da Silva

Solutions Engineer

ACD404

From Side Project to Production: Building a Multi-Tenant SaaS with Kiro

Room: Tech Stage | Time: 16:00

What happens when you use an AI coding assistant to architect, implement, and maintain an entire serverless application? In this talk, I share how I built a multi-tenant SaaS platform using Kiro, shipping 27 releases in under a month. I’ll cover steering-driven development, spec-driven features, where AI excels versus where you still need a human brain, real mistakes and how structured context prevents them from recurring, and how to maintain velocity without accumulating tech debt.

Joris Conijn

CTO for AWS

ACD405

AI-Powered API Mocking with MockNest Serverless: An AWS 10,000 AIdeas Winner

Room: Tech Stage | Time: 16:20

MockNest Serverless started as an idea for smarter API mocking and became an AWS 10,000 AIdeas winner. This session shares the engineering journey behind the project: using generative AI to create realistic mocks, testing prompts and agent flows, designing a scalable serverless architecture on AWS, and the challenges of turning an experimental idea into an open-source developer tool through live demos and practical lessons learned.

Elena Van Engelen - Maslova

Independent Lead Software Engineer

ACD406

Deep dive into AWS Lambda Managed Instances

Room: Tech Stage | Time: 16:40

AWS Lambda Managed Instances enables us to run functions on a wide range of EC2 instances while maintaining Lambda’s operational simplicity. In this talk, I will explain what problems do AWS Lambda Managed Instances solve, what limitations do they currently have, and how to create them. I’ll also compare them with the Lambda default compute type and provide some recommendations on when to use each Lambda compute type.

Vadym Kazulkin

Head of Development

ACD407

Building Conversational Agents in Code with AWS Lambda Durable Functions

Room: Tech Stage | Time: 17:00

Building multi-agent AI workflows usually requires complex state machines to handle unpredictable human delays. Building on the core concepts of AWS Lambda Durable Functions, this session dives into applied agent orchestration. We will coordinate Amazon Bedrock conversational agents with messaging APIs for inbound interactions.

Gunnar Grosch

Developer Advocate

ACD408

AI-Powered Chaos: Automating Resilience Discovery & Testing on AWS

Room: Tech Stage | Time: 19:00

A “quick fix” introduced a single-AZ dependency nobody documented. Systems fail because resilience is assumed, not proven. I’ll show how AI agents discover hidden dependencies by scanning both infrastructure AND source code — finding hard-coded endpoints with no retry logic, missing circuit breakers, undocumented API calls. Then I’ll show how AI auto-generates targeted chaos experiments from those findings — testing YOUR actual failure modes, not generic disruptions. Leave ready to find and validate your own hidden risks in hours, not weeks.

Medha Shree

Solutions Architect at Amazon Web Services

ACD409

Making Agents Boring: Agent Platform Engineering and the Paved Road for Agent Fleets

Room: Tech Stage | Time: 19:20

Boring is the goal. We made Kubernetes boring with templates, pipelines, IAM and rollbacks - agents have none of that yet, so every new one relaunches the platform work. This session lays out a three-plane model for agent platforms: a paved road to build on, AgentOps to run on, and a shared trust plane for identity, gateways and cost. Concrete architecture, honest gaps, patterns you can copy on Monday.

Rajan Bhave

AI Practice Lead

ACD410

Beyond the Demo: Testing Production AI Applications on AWS

Room: Tech Stage | Time: 19:40

AI applications are easy to demo but much harder to trust in production. Prompt changes, model updates, tool calls, streaming responses and provider failures can all introduce unexpected behaviour. This session demonstrates a practical testing strategy for AI applications on AWS using Amazon Bedrock. Through live demos, you’ll see deterministic graph tests, and eval-driven development, giving you a layered approach to testing AI with confidence.

Elena Van Engelen - Maslova

Independent Lead Software Engineer

ACD411

RDS authentication: fast, effortless, cheap — pick two

Room: Tech Stage | Time: 20:00

Every RDS connection must prove who it is, and each authentication method makes you to choose two out of three: fast, effortless, or cheap. IAM-based authentication looks like the enlightened pick: no secret to rotate, nothing extra to pay for. Then p99 latency hits the ceiling while the database sits idle. This is the unfiltered story of chasing multi-second latencies through Aurora Serverless, a test lab and RDS Proxy (with real graphs) and why RDS IAM auth never lands on “fast” for free.

Yevhenii "Yev" Dytyniuk

Cloud Consultant/Engineer

Rooms

Tech Stage

Sessions