LLM Engineering with C#

A C# / .NET companion project to Ed Donner's LLM Engineering course (originally taught in Python).

The goal of this repo is to re-implement each exercise from the course in C#, comparing multiple LLM providers side by side — starting with OpenAI and Anthropic (Claude) — while practicing clean, extensible design patterns along the way.

Why this exists

The original course is built around Python and the OpenAI API. This repo exists to:

• Practice applying the same LLM engineering concepts in C# / .NET
• Compare OpenAI and Anthropic (and potentially other providers) using a consistent, swappable interface
• Reinforce good software design (e.g. the Strategy pattern) while working with real LLM APIs

Tech stack

• .NET 8+
• OpenAI — official OpenAI .NET SDK
• Anthropic — official Anthropic .NET SDK (currently in beta)
• Ollama — runs local open-weight models (e.g. llama3.2) behind Ollama's OpenAI-compatible endpoint, reusing the OpenAI SDK with a localhost base URL
• Environment variables (OPENAI_API_KEY, ANTHROPIC_API_KEY) for API key management; Ollama needs no real key since it runs locally

Getting started

Prerequisites

• Visual Studio 2022 (or any .NET 8+ compatible IDE)
• An OpenAI API key (platform.openai.com)
• An Anthropic API key (console.anthropic.com)
• Ollama installed locally with a model pulled, e.g.:

  ollama pull llama3.2

Setup

1. Clone the repo:

   git clone https://github.com/RizwanRumi/llm_engineering_with_c_sharp.git

2. Set your API keys as environment variables (PowerShell):

   setx OPENAI_API_KEY "your_openai_key_here"
   setx ANTHROPIC_API_KEY "your_anthropic_key_here"

   Restart Visual Studio after running setx — environment variables set this way only apply to new processes.

   Ollama doesn't check the API key, so no ANTHROPIC/OPENAI-style setup is needed for it — just have ollama serve running with a model pulled.

3. Open the solution in Visual Studio 2022 and restore NuGet packages.

4. Run the console app (Ctrl+F5).

Exercises

Week	Exercise	Description	Status
Week 1, Day 1	Meeting Notes Summarizer	Turns raw, informal meeting notes into a structured markdown summary (Key Decisions / Action Items / Open Questions). Implemented for OpenAI and Claude using the Strategy pattern, letting the user pick a provider at runtime.	✅
Week 1, Day 2	Meeting Notes Summarizer + Ollama provider	Same summarizer as Day 1, with Ollama added as a third interchangeable provider alongside OpenAI and Claude.	✅

Related

• Python version of the course exercises: week1/community-contributions/rizwan_rumi (PR-based contributions to the original course repo, organized per week)
• Original course repo: ed-donner/llm_engineering

License

For personal learning purposes.