Building a research career in public

The short version: I spent seven years doing research and then shipping machine learning in production. I am now moving into AI safety full-time. I am early in this field, and rather than wait until I have the credentials to look the part, I am going to build in public — sharing the research, the reading, and the open questions as I go.

Where I'm coming from

My path is not the standard one. I did a PhD, led research projects, and published — but in information systems, not machine learning. Then I spent three years as a hands-on ML engineer and engineering lead, taking models from a notebook to production: fine-tuning, multimodal pipelines, inference optimization, MLOps, the unglamorous parts that decide whether a model actually works for real users at scale.

Somewhere in those three years the question that mattered to me changed. It stopped being can we ship this and became what happens when systems like this are everywhere, and the safety properties we tested in the lab don't survive contact with the real world. That question is why I'm making this move.

The bet

The honest framing of my position: I am not a frontier mechanistic-interpretability prodigy, and I'm not going to pretend to be one. What I have is a combination that's genuinely rare in this field — production ML at scale, real research training, and a year of deliberate upskilling through ARENA, the AI Alignment Research Fellowship, and BlueDot.

The field talks constantly about a shortage of researchers. The shortage I see up close is different: people who have deployed these systems at scale and can speak to how safety actually breaks in production. Most safety methodology is built by people who have never had to keep a model alive in front of real traffic. I have. That's the edge I'm betting on.

What I'm working on

Two threads, one underlying concern.

Open-weight and post-deployment safety. Safety is a property of a deployment, not of a checkpoint. A lab can sign off on the weights it releases; it cannot sign off on the fine-tune, the quantization, or the agent scaffold someone wraps around those weights a week later. Once weights are open, a safety property either survives everything the world does to it, or it doesn't hold at all. I want to build the measurement tooling for that gap.

Compositional misalignment. Alignment is tested on single models; the world deploys compositions — multi-agent orchestrations, tool chains, memory-augmented agents. Alignment does not compose linearly. My HCII 2026 paper documents this empirically for multi-agent LLM systems, and it's the thread I most want to pull on next.

In practice, right now that means: an inoculation-against-model-poisoning project with Safe AI Germany, a mechanistic investigation into why inverse scaling happens, and a steady diet of paper notes and replications.

Why in public

Two reasons, neither of them about personal branding.

First, accountability. Writing a thing down for other people to read forces a level of rigor that private notes never do. Second, the field rewards it. Almost everyone I respect in AI safety earned their standing by doing excellent work in the open — reproducible notebooks, honest write-ups, public reasoning — long before anyone handed them a title.

So this site is the workshop, not the trophy case. Some of what I publish will be wrong, and I'd rather find that out fast. If you're working on any of this — or hiring for it, or funding it — I'd genuinely like to hear from you.