Building software that makes regulated industries move at the speed of consumer products.
Eight years of product management across cryptocurrency, upstream oil & gas, and commercial property insurance. I take complex, regulated systems and rebuild them as products people want to use.
At FM Global, I am modernizing the AIM Customer Portal for FM Approvals and shaping how AI is woven into Loss Prevention and underwriting workflows. At Apache Corporation, I led an internal platform that unified field operations data for production engineers across the Permian Basin. At Binance International, I owned the mobile trading experience for our priority international markets and shipped the company’s IT Security Suite.
I trained as a lawyer first — LL.B and LL.M in Energy Law from the University of Ibadan — before moving into product. The two crafts overlap more than they look: read the regulation, read the system, find the gap, ship the workaround. Regulated industries are where speed and stakes meet, and the product job is to honour both at once.
What I optimize for, in order:
A consistent throughline: ownership of complex platform decisions in regulated environments where stakes are high and trust is earned slowly.
Leading the AIM Customer Portal modernization for FM Approvals and shaping how AI is integrated into Loss Prevention and underwriting workflows. Cross-functional ownership across IT, certification engineering, legal, and information security.
Owned the product roadmap for an internal operational visibility platform serving production engineers and field supervisors. Shipped a unified daily reporting and exception-alerting system across the basin operations org.
Led the international mobile trading experience and the launch of the Binance IT Security Suite. Drove agile transformation that cut time-to-market by 40%, and partnered with Insider Risk Management on a platform-wide risk posture program.
Four case studies. Each is written as a self-contained product narrative — problem, decisions, trade-offs, outcomes. Where company specifics are confidential, examples are clearly labelled representative or illustrative.
A multi-quarter program to defend Binance’s position in twelve priority international markets — cutting onboarding friction, shipping a true power-trader experience, and rebuilding the underlying platform so feature velocity could compound.
Binance International needed to defend its position in fast-growing non-US markets where mobile-first traders were churning to local exchanges with native experiences. Latency, fragmented onboarding, and inconsistent localization were eroding both new acquisition and active-trader retention. Leadership had a clear strategic mandate: be the default mobile exchange in our top twelve international markets.
The legacy mobile app was a single-codebase port that prioritized parity with web over mobile-native ergonomics. Time-to-first-trade for new users in Tier-2 markets averaged roughly seven days. Power-trader features were buried, and high-frequency users — who drove disproportionate volume — were defecting to lighter-weight competitors. Technical debt slowed feature shipping to a six-week release cycle.
End-to-end product owner for the international mobile experience. Defined product strategy with the GM for International, owned the backlog, partnered daily with engineering managers across three pods, and was accountable to the Product VP for shipping and outcomes.
Eight weeks of user interviews with 40+ traders across five markets, quantitative funnel analysis on the existing app, and a competitor study covering three regional exchanges. Output: an opportunity map and segment hypotheses.
PRDs for three workstreams — onboarding, pro mode, discovery surface — reviewed with engineering, design, compliance, and risk. Output: signed-off PRDs and a dependency map.
Two-week sprints across three pods. A weekly trade-off review with engineering leads handled scope drift transparently rather than letting it accumulate in silence.
Phased rollout to two markets first, instrumented heavily, then waterfall expansion to the remaining ten as confidence accumulated. Each market rollout was paired with a regional comms plan.
Monthly KPI review with the GM and a quarterly business review with the executive team. Standing “what we got wrong” agenda item to keep learning visible.
A RICE-weighted backlog scored on (a) market-segment LTV impact, (b) regulatory exposure, and (c) engineering cost. Anything below the cut-line moved to a “next horizon” document. I maintained a visible kill list — features explicitly chosen not to build — to keep stakeholders calibrated and to convert “no” from a refusal into a shared decision.
Engineering pushed back — we’d lose three months of feature velocity to platform work. I held the line because compounding velocity afterward was the point. Backed up the call with an explicit ROI memo to the VP.
A heavier ML approach was tabled. The data wasn’t yet defensible and the regulatory cost in some jurisdictions was high. We shipped a simpler localized-signal engine and committed to revisit on a calendar.
Compliance and product agreed: surface complexity is itself a risk signal. Pro mode was opt-in only and gated behind a behavioral threshold.
Velocity is a strategic asset, but only if leadership owns the trade against capability for it. The platform investment landed because the trade was made visible to the business in the business’s own language.
An internal product that replaced the manual morning-report ritual with a single trusted source of truth, exception-based alerting, and the data foundation for downstream forecasting and AI-assisted optimization.
Upstream field operations ran across a patchwork of SCADA telemetry, third-party well-monitoring vendors, drilling reports, and spreadsheets. Engineers and field supervisors were spending hours each morning stitching together a daily picture of basin performance that should have taken minutes. Leadership wanted a single internal product for operational visibility and a more consistent way to push exception-based attention to the right person.
The morning report production engineers compiled was a manual, error-prone process with no audit trail and no ability to drill into root cause. Decisions about well intervention and chemical treatment were being made on stale or inconsistent data. Annualized, the cost of operational latency and missed interventions across a small subset of wells was meaningful enough to justify internal platform investment.
Owned the product from charter through MVP launch. Defined the discovery scope with the VP Operations, ran requirements and design sprints with field engineers, partnered with the data engineering lead on architecture trade-offs, and managed the launch and adoption plan.
A value-vs-effort matrix scored against the VP’s three stated outcomes. The daily report builder went first because it was the highest-frequency, highest-pain user task. Alerting and analytics were held back so we could earn real adoption before adding features. Forecasting was sequenced last — a deliberate choice against the more glamorous capability.
Faster MVP, but coupled us to the warehouse team’s roadmap. Mitigated by negotiating a shared SLA and a quarterly review.
Forecasting was the more glamorous capability; exception alerting was the higher-utility baseline. Held the line on sequencing.
Field teams asked for it. Deferred to a post-MVP phase to avoid splitting engineering effort and to validate desktop adoption first. Documented with an explicit commit to revisit.
Internal platforms rarely fail on technology. They fail on adoption — which is downstream of whether the executive sponsor describes the problem in the same words the user does.
A multi-phase migration of the customer-facing surface, identity layer, and certification workflow APIs — delivered with no certification-window disruption on a cohort-based cutover plan.
The Account Information Management (AIM) Customer Portal was the primary digital touchpoint for FM Approvals’ certification customers — manufacturers and product owners who relied on the portal to manage approval applications, listings, and account information. The legacy system had accumulated more than a decade of patches: expensive to maintain, awkward to use, and a bottleneck for new product capability.
The legacy portal couldn’t support modern compliance workflows, single sign-on at scale, or the data integrations FM Approvals needed to streamline certification turnaround. Customer support volume was meaningful and rising. Internally, every new approval product required disproportionate engineering effort to plumb through the legacy stack.
Led the program end-to-end as the product manager. Owned the migration plan, requirements catalog, cutover strategy, and customer communications plan. Reported into FM Approvals product leadership and was accountable to the divisional executive sponsor.
Inventoried every legacy feature, scored each on usage frequency × business criticality × migration complexity, then produced an explicit rebuild / rebuild later / sunset decision for each. Sunset decisions were socialized with account management before any customer was migrated.
Feature catalog, signed by FM Approvals leadership before customer migration. Counts representative.Six-week feature audit with usage telemetry and account-manager interviews.
Working sessions across FM Approvals leadership, certification engineering, and IT to socialize the rebuild/sunset list and resolve disputes.
Master PRD covering customer-facing surface, identity, and workflow APIs.
Cut into shippable increments mapped to certification customer journeys.
Customer cohorts sequenced by certification activity. Low-activity first; high-volume manufacturers last.
Decommissioned the legacy portal after a verified parallel-run window.
Slower overall, but eliminated the risk of disrupting active certifications during peak windows.
Took longer up front but meant the new workflow surface launched against a clean identity model.
Each mapped to either a redesigned equivalent or a documented removal with stakeholder sign-off. Prevented dragging legacy debt into the new system.
Dedicated workstream for customer-facing communications and training. Reduced the support spike during cutover.
Migrations demand all the craft of a new product launch with one extra constraint: the product is already in customers’ hands and they have built workflows around its quirks. The platform was the visible product; the negotiation was the actual work.
A draft-and-review tool that helps Loss Prevention engineers turn site visit data into prioritized recommendations and underwriting-ready summaries — designed to be additive to engineering judgment, auditable, and governable from day one.
Loss Prevention engineers conduct site visits to identify and quantify property risks at insured locations. Their reports flow into underwriting, account management, and customers themselves. Synthesis was largely manual: engineers spent significant time turning observations into prioritized recommendations and into language tailored to different downstream readers.
The strategic question: could a careful, AI-assisted product reduce the synthesis burden, surface cross-account patterns engineers might miss, and produce more consistent outputs — without compromising the engineering judgment that is the company’s competitive moat?
The product had to be additive to engineering judgment, not a replacement; it had to be auditable; and it had to satisfy a high bar for explainability and data governance.
Product manager for the AI-assisted insights workstream. Owned the product brief, partnered with the data science team on model selection and evaluation, drove governance and explainability requirements with legal and infosec, and managed the rollout to the engineering field organization.
A second module surfaces potential cross-account risk patterns for human review — engineer-in-loop, never auto-routed to underwriting.
A value-vs-risk matrix instead of value-vs-effort. Synthesis assistance scored high on value, moderate on risk — manageable with a draft-and-review pattern. Autonomous report generation scored high on both value and risk and was explicitly out of scope for v1. Cross-account pattern surfacing scored high on value and moderate on risk — included with human-in-loop confirmation.
A choice for trust and adoption over apparent efficiency. Defended this in front of executives who initially wanted the more aggressive option.
No flag goes to underwriting without an engineer’s confirmation. Reduces false-positive risk; slows surfacing slightly.
A controversial choice — edit distance can be noisy. We used it as a leading indicator of model quality; downstream underwriter satisfaction was the lagging indicator.
Autonomous generation, risk-rating outputs, and pricing recommendations were all explicitly out of scope. Reviewed and signed by legal and the executive sponsor.
The discipline in AI-assisted internal products is to ship less capability, more carefully, with the judgment of the people who own the outcome inside the loop. The pattern is portable; the trust is what has to be earned.
Six pillars I bring to every team I work with. The patterns are durable across insurance, fintech, and energy — only the surface changes.
A scored backlog — RICE or value-vs-effort, sized to the situation — feeds a visible roadmap with explicit kill-list items. Prioritization is a stakeholder ritual, not a private calculation. I make trade-offs visible so the business owns them with me.
Three discipline points. (a) A single source of truth for status and decisions. (b) Pre-reads before every executive review — never surprises. (c) An explicit forum for disagreement before commitments are made; disagreement after commitment is more expensive than disagreement before.
Time-boxed discovery, written hypotheses, evidence-driven rejection or commitment. I would rather kill a question than carry it. Where evidence is genuinely unavailable, I make the assumption explicit and revisit on a calendar.
Frame every trade-off as a comparison, not a choice. The question is never “should we ship X” — it is “should we ship X instead of Y.” Naming the alternative forces real prioritization and surfaces the second-order cost.
Discovery and delivery run in parallel, not in sequence. Discovery is continuous, fed by a backlog of opportunity hypotheses. Delivery operates on a tighter cadence with clear acceptance criteria and instrumentation built into the spec, not bolted on.
Every initiative has a leading indicator instrumented into the build and a lagging business-value indicator reviewed quarterly. Vanity metrics are explicitly excluded. Quarterly reviews include a standing “what we got wrong” agenda item.
“The hardest product decisions are not what to build. They are what to stop building, and how to make that decision visible enough that the business owns it with you.”
— OA, on running discovery in regulated environments
I’m open to senior product roles where the problems are messy, the customers are real, and the bar for craft is high. Insurance, fintech, energy, AI-assisted internal tools — or any environment where regulation and product velocity have to be reconciled.