About
We build agents from a team's expertise. Between us, thirty-five years of building products people depend on and shipping systems that stay up.
Where the founders have built
Founder
Fifteen years building AI products at the frontier: Meta’s first AI assistant for advertisers, the first multimodal Alexa at Amazon, the first conversational AI for classrooms at Merlyn Mind, where he was Head of Product. He works on judgment capture: getting expertise out of people’s heads and into systems that keep it. Wharton MBA, industrial engineer.
Founder
Twenty years shipping production systems, at Electronic Arts and then Gusto. He built a production platform for weekly mine data: deterministic ingestion, hybrid retrieval, and a Spanish-language assistant where the model picks the question and the server computes every number. Master’s in Computer Science, University of British Columbia.
Where the method comes from
The knowledge acquisition bottleneck stalled expert systems in the 1980s: the rules lived in people and nobody could write them all down. Large language models removed the formalization step and moved the bottleneck to elicitation, a field with forty years of research behind it. That is the work: getting what your experts know out of their heads and into artifacts the company owns.
Next
Your people keep the judgment. Lucum turns their rules and exceptions into a tool the company owns.