Now¶
A snapshot rather than a homepage that pretends to be frozen in time.
Working on¶
A knowledge platform for a global CPG manufacturer wrapped its prototype phase this month. The scale-up case is with the client and the decision is outstanding. Whatever comes back, the interesting part is done: the platform produced a measured quality baseline and an honest account of the distance to production. Written up in Work.
This site, rebuilt from a CV-with-navigation into something closer to a working knowledge base. The engineering was the easy half; deciding what could honestly be published from client engagements took longer than building the pipeline that publishes it.
A whitepaper on semantic architecture for agentic systems, in a sourcing pass before editorial review. Details in Writing.
Thinking about¶
When a semantic layer needs an owner rather than a document. Every knowledge platform I have built eventually stalls in the same place: a definition nobody is accountable for. I want to know the smallest governance structure that actually prevents drift — because the large ones do not get adopted, and the absent ones do not work.
What an evaluation gate looks like for an agentic system in an enterprise. Competency questions and LLM-as-judge scoring get you a baseline cheaply. Turning that baseline into a release gate a business will respect — one that can actually block a launch — is where I have not seen a convincing answer, mine included.
How much of delivery management survives AI-assisted engineering. I now build roughly as much as I coordinate, and that changed inside eighteen months. Some of what I used to do as a delivery manager was coordination overhead that no longer exists. Some of it was judgement that matters more than ever. I do not yet know the ratio.
Reading and testing¶
Primary literature on retrieval evaluation and semantic layer governance, mostly — and running the patterns from Methods against situations they were not derived from, which is the only way to find out whether they generalise or whether I got lucky once.
What comes next¶
Priorities are the whitepaper and making the professional profile match the work — both are about distributing evidence that already exists rather than producing more of it. The full sequence, with what blocks what, is on the roadmap.
Open to¶
Conversations about knowledge-system architecture, particularly early ones — when the question is still what to build and how it will be governed, rather than how to wire it. That is where I am most useful and, frankly, most interested.