Building Production GenAI on AWS

A first GenAI prototype is impressive and costs nothing in a demo. The trouble starts with the questions of whether the answers are good enough, what it costs under real traffic, and how to notice a quality drop before users report it. This workshop takes you from a prototype to a GenAI solution that survives production — on AWS services.

Who it is for

A workshop for teams that want to build GenAI on AWS in a production-ready way — with a focus on reliability and cost. It is most useful where a prototype with a language model already exists and you have to decide what comes next so it reaches production without unpleasant surprises.

How we run it

Each module pairs a short introduction with an exercise on a live AWS environment — you build, you do not just listen. You work on a realistic case grown throughout the workshop: from model selection in Amazon Bedrock, through a knowledge base and an agent, to evaluation, observability and cost control.

What you take away

• The ability to choose a model in Amazon Bedrock deliberately, weighing quality, latency and cost
• RAG and agents built with guardrails, identity and observability (AgentCore) from the start
• An evaluation that measures answer quality in numbers, not impressions
• Observability that catches quality regression before a user does
• The ability to model and control the cost of AI workloads (AI FinOps) at production scale