AI Engineer
← All organizations

Cloud computing and AI infrastructure

Amazon Web Services (AWS)

Amazon Web Services (AWS) provides cloud computing and AI services for startups and enterprises. Its portfolio spans compute, databases, serverless applications and managed AI models. Amazon Bedrock gives developers access to foundation models from multiple providers with governance and security controls, while Bedrock AgentCore supplies infrastructure to deploy and operate AI agents. AWS Lambda lets developers run serverless workloads without managing the underlying infrastructure.

Andy Jassy founded AWS within Amazon in 2006 and led it before becoming Amazon’s CEO in 2021; Matt Garman is the current AWS CEO. Its engineering contributions include Firecracker, introduced in 2018: open source virtualization built on KVM that combines virtual-machine isolation with lightweight microVMs. Designed for container and function workloads, it was already powering Lambda and AWS Fargate at launch, addressing the overhead of isolating serverless workloads.

AWS combines managed services with infrastructure spanning 124 Availability Zones across 39 geographic Regions as of August 2026. Its AI offerings serve an established customer base: Amazon reported hundreds of thousands of Bedrock customers in its Q2 2026 results, alongside an AWS $169 billion annualized revenue run rate. Its hardware work also includes custom silicon for general-purpose computing and AI workloads.

aws.amazon.com ↗

16 talks

Newest first

15 speakers at AIE

Affiliations reflect their AIE appearances, not necessarily current employment.

Justin MullerApplied AI architectOlivier LeplusAntje BarthMember of Technical StaffDu’An LightfootSenior AI engineerBanjo ObayomiSoftware engineerDerek BinghamSenior developer advocateMani KhanujaPrincipal Generative AI Solutions ArchitectChintan AgrawalAmazon Web Services solutions architect whose work spans conversational AIDaniel WirjoSolutions architectMike ChambersBrisbane-based Senior Developer Advocate for Generative AISuman DebnathRohit TalluriCo-founder and chief executive of Primitive LabsElizabeth Fuentes LeoneDeveloper AdvocateVikash AgrawalSoftware development managerLinda

Start here

  1. Introducing Strands Agents, an Open Source AI Agents SDK

    Start with Suman Debnath's demonstration to learn how built-in tools, alternative model providers, and an MCP server fit together in an agent that generates mathematical animations.

    Suman DebnathAI Engineer World's Fair 2025

  2. 7 Habits of Highly Effective Generative AI Evaluations

    Use Justin Muller's talk to examine why plausible answers can hide reasoning errors, how test-case coverage affects evaluation, and why results need analysis by category.

    Justin MullerAI Engineer World's Fair 2025

  3. Designing Voice Agents for Real Conversations

    Watch Chintan Agrawal and Daniel Wirjo's Pipecat travel-assistant demonstration to understand the engineering choices behind detecting when a user has finished speaking.

    Chintan Agrawal · Daniel WirjoAI Engineer World's Fair 2026

  4. Building Agents with Amazon Nova Act and MCP — Du’An Lightfoot and Banjo Obayomi

    Du’An Lightfoot and Banjo Obayomi lead a hands-on AWS workshop on building autonomous browser agents with Amazon Nova Act, Model Context Protocol servers, and Strands Agents.

    Du’An Lightfoot · Banjo ObayomiAI Engineer World's Fair 2025

Messages from the stage

Connecting agents to tools

Suman Debnath introduced an SDK organized around models and tools, while Du’An Lightfoot and Banjo Obayomi demonstrated browser automation with Nova Act and MCP. Mike Chambers followed a simple Python agent through deployment concerns.

Finding failures and controlling responses

Justin Muller framed evaluations as a way to identify actionable failures. Mani Khanuja covered retrieval and PII protection, while Elizabeth Fuentes Leone demonstrated semantic tool selection and booking guardrails using simplified tools she cautioned were not production-ready.

Managing conversational timing

Chintan Agrawal and Daniel Wirjo compared voice activity detection, provider endpointing, and local turn classification. Their demonstration connected these choices to streaming pipelines, interruptions, and latency tradeoffs.

Affiliations reflect each recorded session, not necessarily current employment.

Company sources · checked 2026-08-27