BOOK 1 OF 12 · ROYAL CARIBBEAN GROUP (2017 TO PRESENT)

# Cruise AI

Unless you already spoke cruise lingo, planning a trip used to feel like homework: Price: Low-to-High buried great family itineraries under off-season short sailings, keyword search returned generic links, and booking two connecting rooms risked losing one mid-checkout.

Over nine years at Royal Caribbean Group, working alongside brilliant engineers, scientists, and designers, we solved those bottlenecks one production system at a time while I stayed hands-on in TypeScript, evaluation harnesses, and working prototypes. Using geographic cold-start signals (DMA, state, and country) let our 0-to-1 ML ranking model personalize visit one (+13% conversion in A/B tests, $300M/yr across web, mobile, and call center). We followed with Saved Favorites & Price Watch, multi-model support RAG across Claude, GPT, and Llama 3 (40% contact deflection; containment lifted from 49% to 56%), a Databricks feature store behind an omni-channel platform that cut contact-center costs, real-time merchandising data as a service ($80M/yr), and our May 2026 AI answer engine scaling from 42,000 to 110,000+ weekly searches at 95% helpfulness.

- $300M/yr +13% ML lift

- 110K+ AI searches/wk (95% helpful)

- 40% RAG deflection

Next: [Next Voyage](/posts/next-voyage/)

## Say hi or grab PDF

[Contact](/resume/#contact) [Resume](/resume/) [Notes](/notes/)

AI Product & Engineering · Miami

[SKILL.md](/SKILL.md) · [llms.txt](/llms.txt)
