Jose
Vargas

Appointment scheduling assigned customers to whoever was free. Providers with the wrong capability profile got booked anyway, capacity sat idle, and conversion suffered: a cost problem and a revenue problem wearing the same hat.

I defined the solution architecture for a Provider Recommendation Service that matched customers to providers on availability and capability, then drove full-cycle delivery with my engineering team. Once it shipped, I extended it with a performance scoring layer that prioritized providers with higher closing rates and lower return rates.

Routing stopped being a coin flip. Wasted capacity on salaried providers dropped, and the revenue impact was the largest single contribution of my career so far. Two years on, the matching logic is still the foundation: another team built on it this year to launch fully automated voice AI agents that call previously uncalled leads and schedule appointments with no human in the loop. Over the trailing 90 days, that's generated ~$2.0M in net booked revenue, adjusted for expected returns.

Providers were writing outcome notes by hand after every appointment, roughly 5 minutes each across ~180 appointments a day. It was the single biggest time drain outside the appointment itself, and none of that data was structured or usable for anything beyond the note.

I defined the product strategy for an LLM-powered transcription pipeline built on Zoom RTMS, auto-generating outcome notes directly from the appointment.

~3,900 provider hours saved annually, and it created something that didn't exist before: the company's first structured data layer for provider coaching.

The legacy platform wasn't built for how North American providers actually worked. Appointments ran long. Providers self-reported averages in the 60-70 minute range, with a meaningful share pushing past 75, and performance broke down for anyone on a slow connection. It was a platform designed elsewhere, retrofitted for a market it didn't fit.

I owned the US market adaptation of a cross-regional rebuild, applying direct provider relationships and pain-point knowledge to map the new platform to North American workflows, not just porting the old one over.

Average appointment time has roughly halved to 33-36 minutes, and the outlier ceiling dropped from 75 minutes to 45. It's still trending down as we keep shipping.