Building production systems are different from that building a chatbot that works. They need observability, governance, cost controls, tool orchestration, and human approval workflows before they can be trusted in real applications.
In this hands-on workshop, you'll build "JeolAI", a production-inspired AI shopping agent using "Vertex AI", "FastAPI", "React", and "Gemini models".
Together we'll build an agent that can:
- Route requests to different models based on budget policies
- Call multiple backend tools for search, inventory, promotions, cart management, and checkout
- Track tokens, latency, tool calls, and estimated cost for every request
- Visualize agent execution in real time
- Apply budget guardrails that automatically change model behavior as spending increases
- Enforce domain guardrails to reject out-of-scope requests
- Require human approval before completing high-cost operations
- Generate a complete session summary with execution metrics and conversation history
Along the way you'll learn how to structure a production AI application using FastAPI, connect it to Vertex AI, implement deterministic tool calling, and build the operational controls that engineering teams expect in real deployments.
By the end of the workshop, you'll have a complete end-to-end AI agent running locally, understand how modern agent architectures work, and leave with a project you can continue extending after the session.
Prerequisites
- Basic Python knowledge
- Familiarity with REST APIs is helpful
- Please create a FREE Gmail account (gives $300 credits)
- VS Code installed
- A laptop capable of running Python 3.11+
Bring your laptop. This is a build-along workshop, and you'll leave with a working production-style AI agent running on your own machine.
Speaker
Deepa Subramanian
About the Speaker
Deepa Subramanian is a Google Developer Expert (AI & Web) with 15+ years of experience in software engineering across Adobe, Google, Apple, Microsoft, and Mozilla. She is the founder of TeachTech LLC and author of AI Architecture Weekly, where she explores production AI architecture, agentic systems, and engineering best practices.
Location
1981 Chester Dr, Tracy, CA 95376
Parking and Basic Information
The venue has plenty of no-cost parking. If possible, use the south parking lot, closer to Lowell Ave since it is closer to the classrooms within the building.