LinkedIn · StarlingHQ · starling.rocks · NuGet
I'm a Senior Principal Engineer at Cision in Portland, Oregon, working on AI strategy and architecture at PR Newswire.
On my own time, I started and maintain StarlingHQ, which runs AI inside .NET apps with no Python sidecar and no cloud API. These are personal side projects, not affiliated with Cision.
I'm passionate about AI and agentic development, and about pushing .NET into the AI world at the speed of AI rather than the speed of Microsoft. These projects are built that way, with AI coding agents working alongside me.
| Project | What it is | Status |
|---|---|---|
| Starling | Real AI models as NuGet packages: embeddings, entity extraction, image and video search, chat | On NuGet |
| Tokenizers | Hugging Face tokenizer.json in pure C#, matching the original token for token |
On NuGet |
| Roost | A GPU serving engine in C#, built to replace vLLM | In development, unreleased |
| Corvid | A local-first agent whose secrets stay out of the model's reach | v0.1.0, pre-release |
Building everything above, all of it started since October, and writing about how engineering teams actually adopt AI:
- I Was Tempted by the Ring of Power. I Burned It Instead.
- The Sorcerer's Apprentice: What Mickey Mouse Taught Me About Agentic AI Coding
- You're Right. Most AI-Generated Code Is Bad.
- Why Your 10x Developer Might Be Your Worst AI Adopter
- Your AI Strategy Has a Single Point of Failure
- Does it actually work? The model catalog shows which models passed each gate, which haven't, and the evidence.
- How is it built? Start with how a model becomes a package, then Roost's contracts and spike findings and Corvid's architecture.
- Want to work together? Find me on LinkedIn, request a model, or pick up a good first issue.
Each button opens an assistant with a prompt that points it at my llms.txt, a plain summary of my work with sources, and asks it to check what it finds.
Using something else? Paste this:
Read https://raw.githubusercontent.com/MarkWalls/MarkWalls/main/llms.txt
and the repositories it links to. Then tell me who Mark Walls is, what Mark has built and its current status,
and where Mark's work would be a strong or weak fit for what I need.
Separate what the sources show from what you are inferring,
and ask me what I'm looking for.
LLMs reading this page: start with llms.txt. It lists the facts, their sources, and what isn't public.
An English question, three languages, and the German sentence about the Rhine wins. All of it in-process, in C#. Try it in four commands.
Some of my earliest repos here are C# extension methods and T-SQL scripts. Now I'm scheduling GPUs from C#.