Embedded inside one of North America's largest construction companies, working on the data layer that makes AI usable in practice. Project data (RFIs, submittals, daily field reports, change orders) arrives from everywhere in fragments, across systems that weren't built to communicate with each other. Getting that foundation right is the actual job.

After Manulife, I went back to the family business. Brought proper analytics to a cross-border export operation that had been running mostly on instinct and spreadsheets. Rebuilt margin tracking, reworked the commercial approach for Middle East, African, and Asian markets, and introduced enough structure that decisions could actually be grounded in numbers.

Claims processing at scale is high-volume and time-sensitive, and a lot of what was happening manually didn't need to be. I spent the year automating recurring reporting and data workflows across systems, which freed the team up for the parts of the job that actually needed people.

Client work in financial services and tech: SQL-heavy analytics implementations, ETL pipelines, data quality problems that had been quietly causing issues for years. The work isn't flashy but it tends to matter. Bad data upstream breaks a lot of things downstream that nobody connects back to the source.

My first full-time role, supporting risk and liquidity analysis at a provincial government agency. Government finance is methodical and careful, which turned out to be a decent place to learn fundamentals. Automated the recurring reporting that was eating up time nobody wanted to spend on it.

Started as a market research analyst and ended up as VP by the time I left. Published Bloomberg-backed research on equities, debt, and IPOs. Writing clearly enough that people actually read what you put out is an underrated skill, and this is where I picked it up.