I build backend systems that take action.

A white-label platform for agencies and MSPs. Build an AI workflow once, deploy it to every client from one cockpit, and resell it under your own brand. Each client gets an isolated workspace with its own encryption boundary, audit trail, and human-in-the-loop approvals.

Otto is the part I like most: an LLM agent that drafts an executable workflow from a plain-language description. Workflows are data, a JSON node graph run by a sandboxed interpreter, which is what makes "patch it once, push it to all clients" possible.

A fact checker that listens to a live conversation and stays quiet unless something stated as fact doesn't hold up. Every flag carries the disputed quote, an accuracy score, the correction, linked sources, and a confidence level. Built with debates, podcasts, and interviews in mind.

Short audio segments go to Deepgram for transcription, a fast Claude pass extracts check-worthy claims, and a second pass researches each one with web search and renders a verdict. The same engine is exposed over REST and as an MCP server.