I build AI agents that run themselves and answer to a person, and I design the screens people meet them on. Two dozen live.

For seven months I was the content engineer at a security company. I built hundreds of phishing-training emails, landing pages, and login screens, in four difficulty tiers, for more than eight markets. Every one had to follow a written ruleset about which red flags belong in which tier. Doing that by hand does not scale, so I turned it into a pipeline. Claude Code skills hold the rules. An orchestrator runs the eight steps in order. Twelve sub-agents do the writing that needs judgment. Python packages the result the same way every time.