eXp Enterprise: piloting the system of record
Product Manager — AI & Enterprise Platforms · May 2023 – May 2026
What it was
eXp Enterprise was the source of truth for agent data at eXp: financial, demographic, and licence. When an agent joined the company, everything they reported on the join form eventually ended up there, including the most important part, their real estate licence. It held a record for every agent in the U.S. and Canada, built in Mendix on a Kafka and AWS backbone, and dozens of systems consumed its data downstream, including single sign-on and the company's data platform.
Enterprise wasn't just a communicator. It was an application people did work in, and an agent needed their licence transferred there in order to get paid. When Enterprise was down, it was an entire company event.
What I did
I was in charge of the direction of Enterprise. A lot of stakeholders to answer to, a lot of different winds blowing, but I was captain of the ship. All the problems came to me, and all the questions came to me, as we moved forward.
- Owned the direction of a ten-year-old platform. It didn't need a sharp pivot; it needed someone to stop the ship from falling apart while pointing it somewhere. Roadmap three months out, a real intake process, and prioritization that held under executive pressure.
- Ran the permission structure end to end: SOC 2 compliant, FINTRAC compliant for Canada, aligned with the IAM team's job-duty matrices, and auditable every time a role changed.
- Owned the Kafka side: topics and schemas, plus the contract negotiations with every consuming application that depended on them.
- Financial operations: W-2 collection, pay plans that varied by agent situation, and the licence-to-payment flows the business ran on.
- Led without authority: after budget cuts the team was me, a portfolio manager, a QA lead, and seven or eight developers. Nobody reported to me, but I got answers for people, gave guidance including technical guidance, and sat in on code reviews.
- Picked my battles with the platform: Mendix defaults were easy wins; anything unique cost technical debt. Knowing which compromises to take, especially on UI, was product judgment as much as engineering.
- Used AI as institutional memory: ten years of history scattered across Jira, Confluence, Coda, and Google Drive. AI retrieval became essential for impact analysis, for finding how things connected, and for surfacing what mattered in a feature presentation.
Outcome
Three years of stability and direction for the platform the whole company depended on, with a compliance posture that survived audits and a downstream ecosystem that kept its contracts. The last feature I delivered on it, Custom Agent Flags, closed one of its longest-standing gaps.
Stack: Mendix · Kafka · AWS · Okta · Jira · Confluence