Own signing up, proving who you are, and the first payment, across four consumer brands and 22+ US states plus Canada.
Moved the ID check inside the sign-up itself, on iPhone, Android and the web, instead of sending people away to finish later. The share who got through rose from 75% to 92% in the oldest market and 47% to 66% where the law requires an ID; the queue of people left waiting fell 38 to 64%.
Relaunched a line of business with 73.5% of people finishing on day one, 27 points above the month before, closing a nine-month gap that opened when the law started asking for an ID.
Took the share of people who do something after their first payment from 1% to 7%, by changing the order of what they see and measuring where they stop.
Leading the move to a new ID-checking company in every market, built so it can be switched back in seconds from a setting. The first market let 10.8% more people through and rescued 15.4% of the checks that used to fail, with nothing else breaking.
Found a serious security hole in how people get back into their account, raised it, shipped the fix, and wrote the plan to close it with every company involved. What it was, in the case files.
Built the measurement that split one big “they finished” number into seven separate reasons people stop. It showed that a large share of what everyone counted as new users was neither new nor human. The volumes and the shares are in the case files.
Redesigned what happens the first time someone opens the iPhone app, and the way it asks permission to send notices. The share who say yes went from 34% to about 50%.
Product Manager, User Acquisition & First-Time Setup
Caesars Digital · Remote
Owned how people arrive, sign up, log in, prove who they are and pay for the first time, across four consumer brands and 20+ North American markets.
Grew monthly first-time payments 18%, with 24.1% more people finishing sign-up and a 4.78-point rise in payments.
Lifted the share who finish the ID check from 52% to 62%, by simplifying the legal and privacy steps with the Legal team and running the new version against the old one on live traffic.
Cut accounts suspended by mistake, because the same ID looked like two people, by 55.6%, and cut the wait for a security code from eight seconds to under one.
Redesigned logging in and getting back into a locked account, end to end: security codes, fingerprint and face, and an easier way to reset a password.
Owned the plan for every outside company that checks identity, fraud, sanctions lists and addresses.
Added 20+ new measurements across sign-up and the life of an account, so we could see what people actually do instead of guessing.
Product Management Consultant (part-time)
Carnival Cruise Lines · Remote, Miami
Led the research and launch of messaging inside the Carnival HUB app, so guests could write to each other and to the ship in real time, built to work on the weak connection you get at sea.
Product Manager, Mobile & Kiosk
Spirit Airlines · Miramar, Florida
Led the redesign of the app and the airport kiosks: checking in, seeing your trip, changing a booking, and buying bags and seats.
Cut the time it takes to check in 40% and sold more extras at the kiosk and in the app.
Built the way to try two versions of a price or an offer on live traffic and keep the one that works.
Earlier product and commerce roles
The Ritz-Carlton Yacht Collection · Spirit Airlines · Office Depot
Launched everything a guest does before sailing: booking, the loyalty programme, payment and checkout.
Grew marketplace revenue from $20K to $65K daily at Office Depot by tuning prices, stock, and how fast things sold.
AI systems I have built
These are mine: I designed them, built them and I run them. Most of what I think about AI comes from that, not from reading about it.
Software that runs work on its own
Cortex / Jarvis agent OS
A fleet of small AI workers and the service that hands out the jobs, decides how much each one may do alone, and stops at a person when it matters. Each gets a narrow brief, its own copy of the work, and tests that must pass before anything goes live.
AI that talks to people
An assistant that qualifies leads, and the exam it has to pass
A live chat assistant that remembers the whole conversation, knows when to hand over to a person, and has to pass more than a hundred tests and ten scripted conversations before each release. Its voice was tuned against roughly a thousand real messages.
Repetitive work, done unattended
Find, score, and write one by one
Finds candidates in many places, ranks them, writes a document tailored to each one and turns it into a PDF. 1,200+ handled so far, behind checks that stop everything when something looks wrong, and a human stop before anything that cannot be undone.
Systems that recommend
A recommendation engine a person approves
A recommendation engine over 815,000 past records. It says how sure it is, refuses to answer on old data, keeps score of whether it was right, explains itself in cards you can read, and can be steered by chat. 235 tests.
Getting found
Search and content, published automatically
Writes and publishes content on client websites in English and Spanish, finds what people are searching for, and keeps the files that search engines and AI assistants read up to date. Every page is measured, and nothing publishes without passing its checks.
Identity and self-service, designing the steps, account management, logging in, account recovery, first-time setup, sign-up, ID checks, security codes, cleaning up fraud.
Technical, Claude API and the Anthropic Agent SDK, handing work to AI agents, ways to grade what they produce, dashboards to steer them, FastAPI, pytest, Next.js, React, Supabase and PostgreSQL, Vercel, Railway, GitHub Actions.
Languages, English and Spanish, both native.
Education
B.A. International Business and Finance, Florida International University, 2016.
My current employer is described by sector rather than by name, and the figures here are conversion and performance rates rather than internal revenue. Everything else matches my LinkedIn profile.
