I make complexity
look effortless.

Partnership Intelligence Platform

Know everyone before the room.

The business problem

Partnership conversations are won before the meeting starts. The research that wins them already exists. It is just never in one place, in one voice, at the moment someone has to walk in and ask.

What I did

Defined the dossier standard: what an executive needs to know about a sponsor or a decision-maker, in what order, at what confidence. Then designed the command center and built it.

Judgment I used to apply deal by deal, turned into a product other people can run without me.

Operations Assistant

An assistant that will tell you when it does not know.

The business problem

On an event day the person who needs the credentialing rule has a radio in one hand and ninety seconds. An assistant that guesses in that moment is worse than no assistant at all.

What I did

Built retrieval over four operations manuals with a fixed answer format: primary source, supporting sources, a confidence rating, the procedural language verbatim. When the manuals do not cover it, it says so instead of inventing policy.

Retrieval, citation and calibrated refusal as product behavior, not as a policy nobody reads.

Running agents the way a show is run

Nothing goes live because it seemed fine.

The business problem

Teams adopt agents faster than they adopt any way of telling whether the agents are behaving. When something does go wrong, there is no record of what was approved, what actually ran, or who decided.

What I did

Stood up and operated an agent-native stack on my own Linux VPS: multi-agent workflows coordinated through Mattermost threads, approvals routed to me before anything executed, scoped secrets, and evidence captured at every stage. Then I drilled it. I planted a failure mid-pipeline to prove the stop actually stops, recorded the correction in my own words, and reran clean from the start with separate artifacts throughout.

Standing an agentic system up, operating it, and proving to someone else whether it is behaving.

Attribution. The underlying framework and its agents are AI Build Lab's work, not mine. The implementation, configuration, operation, evaluation, failure testing and documentation described here are my own.