.NET AI Jumpstart
Go from "we should add AI" to a working agent in your codebase, with the patterns and practices to build the next one yourselves.
- Format
- Instructor-led workshop (in-person or remote)
- Duration
- 2 days
- Level
- Intermediate
Who this is for
- Senior .NET developers and architects
- Engineering leads evaluating AI adoption
- Teams with existing C# codebases targeting AI augmentation
Curriculum
What the program covers
Module 01
Day 1 — The .NET AI Stack
- Microsoft.Extensions.AI: the abstraction layer, IChatClient, and why it matters for testability and provider portability
- Choosing your model: Azure OpenAI, GitHub Models, local via Ollama — provider differences and when they matter
- Prompt engineering for engineers: system prompts, role boundaries, structured output with JSON schema, and the failure modes worth knowing
- Semantic Kernel fundamentals: kernel construction, function registration, plugins vs. services, and the DI model
- Retrieval-Augmented Generation end-to-end: chunking, embedding, Azure AI Search, and measuring retrieval quality
- Lab: build a RAG feature against a real document corpus in your own codebase context
Module 02
Day 2 — Agents and Production
- Microsoft Agent Framework architecture: AIAgent types, ChatClientAgent, tool dispatch, and session state
- Building typed tools: attributes, parameter binding, error handling, and the contracts agents rely on
- MCP server development: exposing internal APIs as MCP tools so any agent host — Copilot Studio, GitHub Copilot, custom — can reach them
- Observability: OpenTelemetry on Microsoft.Extensions.AI, token counting, latency tracking, and what to dashboard
- Evaluation: building an offline eval harness with real inputs, grounding checks, and a CI gate
- Lab: wire an agent with two MCP-backed tools, add an eval test, deploy to Azure Functions
What this isn’t
There are a lot of AI workshops that walk you through an OpenAI SDK quickstart and call it a
day. This isn’t one of them. The Microsoft AI stack has moved significantly in the last
eighteen months — the Agent Framework, the Microsoft.Extensions.AI abstractions, MCP —
and most training material hasn’t caught up.
This jumpstart teaches the patterns your team will still be using two years from now: the DI-native, testable, observable approach to AI components that your architecture and your auditors can actually accept.
What you leave with
By the end of day two, every participant has run both labs against real code and a real Azure environment. You leave with:
- A working RAG implementation in a .NET project you can reference or ship
- A working agent with MCP-backed tools, an eval harness, and Azure Functions deployment
- The mental model for where
Microsoft.Extensions.AI, Semantic Kernel, and the Agent Framework each belong in a production codebase - Lab source code, architecture reference sheets, and a curated resource list
How we run it
Maximum twelve participants — larger than that and the labs lose their value. We run the environment on your Azure tenant or on a provisioned tenant we set up in advance; either way, participants work in the same environment they’ll deploy to. If you have a specific internal system you want to wire up during the MCP lab, bring it — we’ll adapt.
Remote delivery works well for this material; we’ve found two focused days beats a slow-drip series every time.
Prerequisites
- Comfortable with C# and .NET 8+ (generics, async/await, DI)
- Working Azure subscription (trial is fine) with permission to create resources
- Visual Studio 2022 or VS Code with C# Dev Kit installed
- No prior ML or AI background required
Request jumpstart details
We'll send a detailed agenda, prerequisites checklist, and lab environment setup guide.
Typical lead time is two to four weeks from inquiry to confirmed date
Maximum cohort sizes are enforced — labs need room to breathe
In-person (Chicago or your site) and remote delivery both available
Get the details
We'll send a detailed agenda, prerequisites checklist, and lab environment setup guide.
More training