Yu Fong C.

I build and test AI agents for chat and voice. Before launch, I measure how often each one answers correctly, so you'll know what to expect. If yours is already live, I'll find out why it's making mistakes and fix it. I've worked on an hourly contract with a conversational AI company that started as a paid trial. I tested their answers in English and Spanish and built the monitoring and regression checks around their live system. Some recent work: • A voice assistant MVP, delivered in a two-week sprint and rated five stars by the client. • An evaluation of a document AI agent. I scored its answers, traced where it went wrong, and handed over a report the team could act on. They kept coming back. • A freight quoting engine that prices from the carrier's rate card, checks each quote against the carrier's own calculator, and flags any difference. • A support inbox that brings Telegram messages into a helpdesk, so the team answers from one place. How I work: I'm in Taipei, and my evenings overlap with the US Eastern morning. I scope and quote before starting, send written updates as I go, and hand over documented code you can run without me. Tell me what your agent does today and where it slips. I'll reply with how I'd approach it.