Product Director · Fintech · SaaS · AI-Powered Products
10+ years leading product across cybersecurity, fintech, retail media, and banking. I currently own a portfolio of 8+ AI-powered products end to end, from opportunity assessment to commercial launch, and lead multidisciplinary teams that move fast from idea to working prototype.
Selected work
Different projects across my career, spanning AI, SaaS, and fintech, each one an example of taking an ambiguous, high-friction problem and turning it into a scoped, shippable product, plus two earlier case studies in growth and conversion, one in fintech and one in retail.
Automated user access review, with an inference engine that auto-drafts the review model.
One intake triggers the entire onboarding sequence across every contract type.
An AI layer that triages email and Slack into decision cards an executive can swipe through.
Website redesign and CRO for a digital bank, across cards, loans, and insurance.
Catalog intelligence for a small-appliance range: an AI agent workflow to clean, match, and personalize at scale.
How I work
Leadership isn't only roadmaps and org design. I stay hands-on as the Product Owner on the initiatives that need it, close enough to the backlog and the day-to-day to make the calls, while also building the team shape and giving people what they need to move fast.
I currently own the vision, strategy, and roadmap for a portfolio of 8+ AI-powered products (SaaS and on-premise) at EVOCS, including P&L and annual budget for tooling, contractors, and resourcing.
Current portfolio includes: AURA · JML · EVE · PSA · Brain · Spectra · Atlas
I lead a multidisciplinary team of 4-7 direct reports that goes beyond traditional PM roles:
Craft
The shape of my process depends on whether the logic behind the product is already known or has to be learned from data. Explicit business rules get a deterministic model: mapped, agreed, and built. Anything that has to predict from historical behavior gets a probabilistic model: validated in stages before it ever touches a real user.
Used when the rules can be written down: eligibility criteria, contract types, approval thresholds. Same input, same output, every time, and every rule traceable to a decision someone signed off on. This is the framework behind AURA and JML.
Weighted 1–5 scoring across Market opportunity, Strategic fit, Demand evidence, Readiness to ship, Effort, and Risk, feeding a single portfolio-wide prioritization report so every opportunity is judged apples-to-apples.
PMO-owned. Accountability is assigned before a single spec gets written, across Product, Engineering, Tech Writing, and PMO.
Product owns the PRD and wireframe, Tech Writing validates it for clarity, Engineering scopes the high-level architecture and builds a proof of concept, validated directly with the CEO before it scales.
Full PRD + user stories + wireframe, a prioritized backlog, and feedback documentation: the package Engineering needs to build without back-and-forth.
Product and Engineering define UAT use cases together, QA runs the cycle, and every release closes with a go-live ceremony, not a silent deploy.
PRDs, BRDs, Opportunity Assessments, and market/competitive research aren't paperwork on the side. They're versioned artifacts with named owners, tracked the same way code is.
Used when there's no fixed rule to write down: personalization, matching, fraud detection, anything that predicts from historical patterns instead of following a static one. The output is a confidence score, not a certainty, so the process spends more time validating before it ever reaches a real user.
Agree on the real signal to optimize for, engagement and redemption per user, not a proxy like raw clicks, before a single model gets touched.
Confirm the historical data the model needs actually exists, is accessible, and is clean enough to learn from. This is where Engineering gets pulled in first, as a data-access request, not a build.
Ship the simplest rule-based version first, so there's a real number on the board to beat before investing in anything more complex.
Test candidate models against historical data, then run the model live but hidden, comparing its calls to what actually happened, with nothing user-facing yet.
High-confidence predictions apply automatically, anything below the threshold routes to a human reviewer, and the whole thing runs first with a control group before a full rollout.
Unlike a rules engine, a model's accuracy decays as behavior shifts. Performance gets watched against the outcome metric, and the model gets retrained on a cadence, not left to drift.
Track record
The full breakdown, with metrics per role, is in my CV.
Own an 8+ product AI portfolio; led the transition from consulting firm to product-led company.
Global data platform unifying online/offline transactions across 5+ markets; GDPR & PSD2 alignment.
Led the WizInk website redesign and CRO across cards, loans, and insurance, plus product work for Santander, MaxMara, and MediaMarkt within a digital transformation agency.
Led a 30+ dashboard financial analytics suite adopted by 7 business units.
Scaled LatAm's largest digital magazine kiosk; represented the company at Mobile World Congress.
Education & languages
IE Business School (Master in Business Analytics and Big Data) · UCA University (Business Administration). Professor of Introduction to Business Analytics, IE Business School MIM.
Let's talk
Open to Product Director / Head of Product roles in fintech, SaaS, and AI-powered products.