Agent Observability on Google Cloud
This 3-hour Cloud Sparks session is a short, interactive introduction to operating AI agents with confidence on Google Cloud.
In Agent Observability on Google Cloud we focus on getting Gemini-powered agents ready for production, with transparency, auditability and predictable costs. You will see how an agent’s non-deterministic reasoning, the ReAct loop, maps to Cloud Trace spans so you can debug what the agent actually did. You will also look at Logs-Based Security Metrics for governance, and at alerts and custom dashboards in Cloud Monitoring that help you keep cost overruns and quality drift under control.
The session works through presentations, visual walkthroughs and strategic discussions. Along the way you will set up a continuous quality loop with Golden Test Cases and learn to connect technical observability metrics to business KPIs such as cost and ROI.
This course is aimed at AI/ML engineers, data scientists, SRE and DevOps engineers, MLOps engineers, technical leads and anyone involved in running or governing Gemini-powered agentic workflows. Participants should be familiar with foundational machine learning concepts (models versus agents), Google Cloud services and the console, and software development and DevOps/MLOps principles. Experience with the Google Cloud CLI and Agent Platform services is highly beneficial.