I lead AI products. And I ship the first version myself.

Staffing agencies lose hours daily to manual prospecting. Kurrent replaced that with a 10-stage automated intelligence engine: it found companies actively hiring, enriched each with firmographic data, surfaced the decision-makers, and delivered everything as white-label, client-branded intelligence briefings: daily digest, PDF reports, and a full client portal for working the pipeline.

And it was a product, not a script: role-based views gave recruiters their morning leads. And it gave the owner the picture she ran the business on: pipeline value, projected placement fees, the money still on the table.

Architected, engineered, and deployed end-to-end, in two weeks, and used in production on live prospect work. The client portal is open as a demo of the pattern, on sample data.

Delivery for a major account generated enormous operational data: every blocker, dependency, and delay, all of it sitting unused. I saw the opportunity: turn that record into intelligence. I architected and engineered a platform that captured, classified, and attributed every impediment, revealing exactly where time went: internally, with the client, or with their partners.

Recurring blockers were engineered out before they resurfaced, client conversations ran on shared evidence, and delivery decisions were made on data instead of memory. The CTOs of Tucows' three business units selected it as winner of the company's inaugural AI competition.

The demo above is a generic reconstruction of the pattern on fictional data: Jira, Slack and GitHub evidence, classified and attributed, ending in the list of what to fix.

Every high-growth platform eventually meets enterprise expectations. I implemented and led the delivery framework that met them: five gated stages, dual sign-off, artifacts that made what, why, and how legible to anyone in the organization. Delivery for the company's largest account became predictable and on time, and leadership extended the framework across the company.

Product Manager on an organization-wide AI initiative that plugged people directly into the codebase, in plain language. Sales answering technical questions without pulling an engineer. Support diagnosing incidents at the source. Product managers seeing dependencies across systems they didn't own. An early, working answer to what AI looks like when a company's knowledge stops being siloed.

The demo above rebuilds the idea on a fictional platform: pick a role, ask the codebase, and get answers grounded in the code with the system map lighting up what they touch.