- Lead designer on KAYAK's AI work for business travel. Took a booking flow that ran as a fixed sequence of screens and made it respond to what a traveler is actually asking for.
- Defined the behavior spec and response structure for KAYAK's AI chat, and pushed for a minimal experiment before the full build. That test returned +14% booking revenue.
- Worked with the machine learning team on a dashboard that lets travel brands see what a pricing decision would likely do to their results before they commit to it.
- Removed a manual approval queue that sat between partners and changes to their own campaigns, giving back 1.29 days per request across roughly 1,500 requests a year.
- Designed bulk editing so a partner can set up 500 ad placements in one pass, against the 1M+ route and placement combinations the system resolves per request.
- Designed internal A/B testing tools so product managers and engineers could read their own experiment results without filing a request with the data team.
AI Chat in the funnel KAYAK · Conversational search Shipped · 2025–26
Moving KAYAK's AI into the booking funnel. KAYAK's standalone AI site drew traffic but almost no return visits, so I rebuilt it as chat inside the booking flow and tested it with a single-input experiment that lifted booking revenue 14%. Live on kayak.com.