Company truth lived in five places and zero answers. I built the data lake that joins every source — and on top of it, the platform the business is now run from.

What it answers

• Credit: origination, portfolio state, and where applications stall in underwriting.
• Revenue-share models: how every dollar splits and what each structure actually returns.
• Partnerships: performance per partner, revenue concentration, and where it leaks.
• Growth: patterns, seasonality and real-time monitoring by partner and product.

My role

End to end: architecture with engineering, the SQL and metric definitions, the visual design of every view, weekly iteration with founders and sales. It replaced spreadsheets as the company's operating tool — founders open it every morning.

The flagship was a single-purpose advances app on legacy architecture. I led the rebuild from zero into a super app — a 360° turn on what the company offers.

The process

• Market discovery across the fintechs that set today's bar — what earns the valuation, which patterns are table stakes.
• A product thesis of our own, not a lift-and-shift of the old app.
• Prototyped flows with Claude Code and Cursor before committing engineering time.
• New web and mobile shipped as one coherent experience, parity from day one.

Why it compounds

Built on the Revenue Engine's data layer, so product and analytics share one source of truth — every change reads in the same dashboard leadership uses.

Promoted to Group PM in 18 months. Built the product org's metric system and led a group of PMs across separate products.

What I built

• North-star and guardrail KPIs per product, with shared definitions across teams.
• Standardized dashboards — no more meetings where everyone brings their own number.
• Experiment readouts as the default language for roadmap decisions.
• Management and mentoring of PMs, each with their own product and stakeholders.

Cross-functional alignment with engineering, architecture, data, operations and the executive team — in a publicly listed company, where the number you report has consequences.

An AI audit tool running on GPS data. A route-identification model that reconciles trips automatically and flags what doesn't add up.

From cost centre to revenue line

• Trip reconciliation was manual, case by case — nobody could review millions of GPS points.
• The model reconciles automatically and marks only the exceptions worth a human look.
• Framed as billing accuracy, not surveillance — that reframing is what got it adopted.
• Packaged and sold as a B2B service into Europe and Asia through a global client operating in both regions.

The internal platform that moves the operation. 1M+ people transported a day across 11,000+ units monitored in real time — and for the first time, full visibility of every one of them.

What it took

• Field work in yards and control rooms — the data existed, but arrived late, incomplete and over radio.
• Getting operations, finance, CX and leadership to agree on one definition of a completed trip. Hardest product work of the project.
• Real-time telemetry, API-integrated with operations and finance systems.
• Regional rollout, field team first: if the person in the yard at 5 a.m. doesn't use it, it doesn't exist.

Credit products for underbanked users, at the moment the user base was growing faster than the organisation could absorb.

The work

• Risk wanted more filters, growth wanted fewer. The only way through was putting cohort data on the table instead of trading anecdotes.
• Cohort and funnel analysis to pick which onboarding step to fix first — and the evidence to defend the call.
• Changes measured against the previous cohort, never against the blended average.

Reading cohorts instead of averages is the discipline I took to everything after this.