AiLive! 6-Week Course: Mastering Microsoft Foundry: From First Agent to Production
February 4 – March 11, 2027 | Live Virtual Training
Level: Intermediate
From Your First Model Deployment to an Observable and Governed AI Solution
A successful AI demo is only the beginning. How do you choose a model, give an agent the right tools, measure whether it works, and keep it secure as usage grows? This six-week AILive! workshop connects those decisions through practical demonstrations in Microsoft Foundry and VS Code.
Watch an enterprise knowledge and service agent take shape across the series. Start in the Foundry portal, then use Microsoft Agent Framework to illustrate custom behavior and hosted execution. Follow the same solution through grounding, evaluation, monitoring, adversarial testing, and release decisions.
Why This Course Is a Must-Attend
-
1 Make Better Architecture Decisions
Understand when a prompt-based agent is the right fit and when a hosted agent gives you the control you need. Connect model selection, data access, orchestration, identity, and deployment choices to the demands of real business applications.
2 Learn Through Live Demonstrations
Most of each two-hour Thursday workshop is devoted to instructor-led demos in the Foundry portal or VS Code. See the configuration, inspect the code, observe failures, and learn how to diagnose and improve the result.
- Deploy and compare models with attention to quality, latency, capacity, and cost
- Create and update agents, connect enterprise knowledge, and add tools safely
- Use evaluation results and traces to improve behavior before and after release
3 Build Confidence Beyond the Playground
Explore the operational practices that turn promising prototypes into maintainable solutions: repeatable evaluations, OpenTelemetry tracing, production monitoring, access controls, red teaming, and governance. Dedicated Tuesday office hours provide time for questions, troubleshooting, and discussion of your own scenarios.
Course Schedule
Six Thursday instruction sessions, two hours each, plus five Tuesday office-hour sessions, one hour each, within the seminar dates. All sessions are virtual.
A single enterprise agent scenario connects the demonstrations from week to week. Optional independent practice between live sessions helps you apply what you see.
Certification That Stands Out
Upon completing of the full 6-week Mastering Microsoft Foundry program, you’ll earn a certification of completion issued by AILive! & Live! 360 Events. This credential recognizes your practical knowledge of deploying models, building prompt-based and hosted agents, and using Microsoft Foundry with Microsoft Agent Framework to develop enterprise AI solutions.
Demonstrate to employers and stakeholders that your skills extend beyond the initial prototype. You’ll understand how to evaluate agent performance, investigate behavior through telemetry and monitoring, and apply security, governance, and red teaming practices to prepare AI solutions for production.
Build practical expertise. Earn recognition. Advance your work in enterprise AI.
Course Schedule*
Thursday, February 4, 2027 - Week 1 - Instruction Session | 2 hours
Foundry Foundations and Model Deployments
- Navigate Foundry resources, projects, model catalog, playgrounds, and project endpoints
- Select models for reasoning, tool use, structured output, and multimodal requirements
- Compare standard, provisioned, and batch deployment options; understand processing geography and availability
- Manage model versions, quota, token limits, latency, and consumption costs
- Demo: deploy a model, compare responses, and call the endpoint from VS Code
Tuesday, February 9, 2027 - Office Hours | 1 hour
Questions, demo follow-up, and troubleshooting. Bring your own implementation scenarios.
Thursday, February 11, 2027 - Week 2 - Instruction Session | 2 hours
Prompt Agents and Enterprise Knowledge
- Create, configure, test, update, and version a prompt-based agent in the Foundry portal
- Manage instructions, conversations, tool calls, and structured responses
- Ground answers with file search or Azure AI Search; understand embeddings, retrieval, and citations
- Connect functions and Model Context Protocol tools with clear permissions and approval boundaries
- Demo: build a knowledge agent that retrieves evidence and performs a controlled business action
Tuesday, February 16, 2027 - Office Hours | 1 hour
Questions, demo follow-up, and troubleshooting. Bring your own implementation scenarios.
Thursday, February 18, 2027 - Week 3 - Instruction Session | 2 hours
Hosted Agents and Microsoft Agent Framework
- Compare prompt-based and hosted agents: configuration, custom code, runtime control, and operational responsibility
- Build and debug an Agent Framework agent in VS Code using Foundry models and tools
- Explore sessions, state, streaming, multi-agent handoffs, and human approval in a focused workflow
- Package and deploy a hosted agent; examine endpoints, identity, scaling, and lifecycle management
- Demo: extend the workshop agent with custom orchestration and run it in a managed hosted environment
Tuesday, February 23, 2027 - Office Hours | 1 hour
Questions, demo follow-up, and troubleshooting. Bring your own implementation scenarios.
Thursday, February 25, 2027 - Week 4 - Instruction Session | 2 hours
Evaluations and Continuous Quality
- Build a representative test dataset with expected outcomes and challenging cases
- Measure groundedness, relevance, task completion, tool-call accuracy, and safety
- Compare built-in and custom evaluators; calibrate AI-assisted scoring with human review
- Run portal and code-based evaluations; compare prompts, models, and agent versions
- Demo: identify a regression, improve the agent, and define an evaluation gate for release
Tuesday, March 2, 2027 - Office Hours | 1 hour
Questions, demo follow-up, and troubleshooting. Bring your own implementation scenarios.
Thursday, March 4, 2027 - Week 5 - Instruction Session | 2 hours
Telemetry Monitoring and Operational Readiness
- Connect OpenTelemetry traces to Azure Monitor Application Insights
- Follow model calls, tool execution, and workflow steps to investigate failures and latency
- Monitor errors, token consumption, response time, and quality signals; establish alerts
- Protect sensitive prompts and outputs through deliberate telemetry collection and retention
- Demo: diagnose a failing tool, compare trace evidence, and tune the quality-cost-latency balance
Tuesday, March 9, 2027 - Office Hours | 1 hour
Questions, demo follow-up, and troubleshooting. Bring your own implementation scenarios.
Thursday, March 11, 2027 - Week 6 - Instruction Session | 2 hours
Security Governance Red Teaming and Release
- Apply Microsoft Entra identity, managed identities, RBAC, and least-privilege tool access
- Examine network isolation, secrets, data boundaries, and safe handling of retrieved content
- Test prompt injection and unsafe tool use; apply content safety controls and human oversight
- Use Foundry red teaming to assess risks, inspect attack results, and verify mitigations
- Connect Foundry Control Plane governance, policy, auditability, versioning, and rollback to release readiness
- Demo: test and harden the workshop solution, then walk through a practical production-readiness review
Final instruction session and workshop wrap-up on March 11, 2027.
Put Microsoft Foundry to Work
Enroll in Mastering Microsoft Foundry through Live! 360 and learn how to connect model deployment, agent development, and enterprise operations in one practical workflow.
By the End of This Course You Will Know How To
- Select and deploy a model with clear capacity, cost, and data-processing considerations
- Choose between prompt-based and hosted agents and manage their development lifecycle
- Connect an agent to enterprise knowledge and tools using Microsoft Agent Framework where custom code is needed
- Create an evaluation baseline and use traces and monitoring to improve quality and reliability
- Apply identity, safety, red teaming, and governance practices to a realistic enterprise agent
Attendee Workstation Requirements
For attendees who want to reproduce the demonstrations and complete independent practice:
- A Windows or Mac computer with a current browser and access to the Foundry portal at ai.azure.com
- An Azure subscription with permission to create or use Foundry projects, deploy models, and connect supporting services; model quota and regional availability must support the exercises
- VS Code, Git, and a supported Python or .NET SDK compatible with the Microsoft Agent Framework examples used in the workshop
- A container runtime such as Docker for reproducing hosted-agent packaging exercises
- Working knowledge of Azure fundamentals and basic Python or C#; prior experience with generative AI concepts is helpful
- An Azure budget for model inference, evaluations, search, hosting, and telemetry; usage costs vary by configuration and practice activity
Demonstrations emphasize capabilities available at delivery. Preview features will be identified, and specific models, portal experiences, and deployment options may vary by region and subscription.