Uploaded November 2026 | Updated November 2026, 6 hours ago
Can you trust what your AI agent does? In this [DATE] episode of GenerAIly Speaking, hosts Jay Busch and Bharathi Srinivasan answer that question twice with live demos, once for an agent in software and once for an agent with arms, a camera and a voice.
First, Raviteja Sunkavalli shows how to observe an AI agent from its first prototype through production with Amazon CloudWatch Omni. He traces and evaluates an agent in the IDE while he builds it, sends telemetry with Agent Observability (ADOT SDK), walks the application map, triggers a real failure live, follows it from metrics to traces to logs, and asks Omni to investigate in plain language. He closes the loop by adding the failing case to the evaluation dataset so it is caught before the next release.
Next, Jeremy Wallace shows Rad, an agentic AI companion robot he built with Strands Agents, Amazon Bedrock and AWS IoT Greengrass. A Strands agent runs inside ROS 2, reasons and sees with Amazon Bedrock, speaks with Amazon Polly, remembers people, and asks before it moves while a digital twin previews the motion. You also see how AWS IoT Greengrass deploys it to the robot with short-lived credentials, so no AWS keys live on the robot, and how an AI coding agent helped design everything from the code to the 3D-printed body.
Explore the Amazon CloudWatch Omni interactive demos aws-cloudops.storylane.io/hub/cw-omni
Try Amazon CloudWatch Omni in the Playground https://cloudwatch-omni.global.app.aws/home/?playground
Get started with the Amazon CloudWatch Omni workshop https://catalog.workshops.aws/observability/en-US/cloudwatch-omni
Build your first agent with Strands Agents strandsagents.com
Get started with AWS IoT Greengrass docs.aws.amazon.com/greengrass/v2/developerguide/getting-started.html
Learn more about Amazon Bedrock aws.amazon.com/bedrock
Tune in to watch GenerAIly Speaking every other Wednesday on Twitch, or right here on YouTube! #AWS #AmazonWebServices #GenerativeAI
Can you trust what your AI agent does? In this [DATE] episode of GenerAIly Speaking, hosts Jay Busch and Bharathi Srinivasan answer that question twice with live demos, once for an agent in software and once for an agent with arms, a camera and a voice.
First, Raviteja Sunkavalli shows how to observe an AI agent from its first prototype through production with Amazon CloudWatch Omni. He traces and evaluates an agent in the IDE while he builds it, sends telemetry with Agent Observability (ADOT SDK), walks the application map, triggers a real failure live, follows it from metrics to traces to logs, and asks Omni to investigate in plain language. He closes the loop by adding the failing case to the evaluation dataset so it is caught before the next release.
Next, Jeremy Wallace shows Rad, an agentic AI companion robot he built with Strands Agents, Amazon Bedrock and AWS IoT Greengrass. A Strands agent runs inside ROS 2, reasons and sees with Amazon Bedrock, speaks with Amazon Polly, remembers people, and asks before it moves while a digital twin previews the motion. You also see how AWS IoT Greengrass deploys it to the robot with short-lived credentials, so no AWS keys live on the robot, and how an AI coding agent helped design everything from the code to the 3D-printed body.
Explore the Amazon CloudWatch Omni interactive demos aws-cloudops.storylane.io/hub/cw-omni
Try Amazon CloudWatch Omni in the Playground https://cloudwatch-omni.global.app.aws/home/?playground
Get started with the Amazon CloudWatch Omni workshop https://catalog.workshops.aws/observability/en-US/cloudwatch-omni
Build your first agent with Strands Agents strandsagents.com
Get started with AWS IoT Greengrass docs.aws.amazon.com/greengrass/v2/developerguide/getting-started.html
Learn more about Amazon Bedrock aws.amazon.com/bedrock
Tune in to watch GenerAIly Speaking every other Wednesday on Twitch, or right here on YouTube! #AWS #AmazonWebServices #GenerativeAI










