Building
What I'm building toward
AI systems for institutional work that currently waits on a small number of experts.
Most institutions don’t have a data problem or a model problem. They have a judgement problem: the decisions that matter sit with a few experienced people, and the rest of the work waits on them.
I’m building software for that situation — systems that capture how experts review, decide and sign off, and apply that consistently inside the workflow, in a form an institution can govern.
The work is at the prototype stage. I’ll write about the problems here as I work through them.
01
Start from the decision
Model the judgement an expert applies, not just the documents they read.
02
Keep the expert in the loop
Make expert review faster and better-informed, never optional.
03
Build for the audit, not the demo
Every output should trace back to its sources and the rules that produced it.
Smaller experiments I've shipped along the way live in the Lab.
Get in touch
Working on similar problems?
If you're an operator, investor or institution thinking about where expert judgement and software meet, I'm glad to compare notes.