Who builds this.

Embedded AI leader across six industries: energy, insurance, distribution, legal, enterprise security, and workplace safety. Systems running in production on three clouds, plus two products of his own. Also the author of the list of things that did not work, which is published here on purpose.

The work starts with joining the team: learning every name, attending every standup, understanding the business before touching the data. This isn't a philosophy. It's the only approach that consistently ships production systems.

Based in Portland, OR. Available for embedded engagements nationwide.

43,000 lines of Python, tested by another 36,000 lines of tests. 2,827 of them pass. Edwin is our internal AI research system: a living memory architecture that verifies what it stores, challenges what it believes, and catches 95% of false information before it enters memory. A provisional patent was filed in February 2026 (USPTO 63/991,739), with 10 claims validated against 35 experiments. Two of those experiments disproved what we expected, and the filing says so.

Edwin uses multi-model adversarial verification, where AI models argue with each other to surface the truth. It runs agentic pipelines, maintains cryptographic proof of knowledge state, and continuously consolidates its understanding through counterfactual challenges.

Everything we learn building Edwin flows directly into client work. The same verification discipline that powers our R&D powers your data warehouse.