12+ years in product and growth, seven of them at Berlitz, where I now own the funnel from first search to activation, B2C self-serve and B2B enterprise. I also founded Magic Pulpo, a growth agency and product studio, and build with AI every day.
Replaced per-market builds with one CMS-driven platform: right-to-left Arabic at component level, course pages marketing publishes without an engineering ticket, and Salesforce lead routing by locale and course interest.
Started with one market, a lightweight checkout, real payment rails and Salesforce carrying every order into the CRM sales already lived in. Then scaled the pattern into one commerce architecture across Canada, India, Japan, Germany, the US and the UK. High-intent buyers self-serve, complex cases route to consultants.
One seasonal playbook run across Canada, the UK and beyond: paid search, landing pages and email journeys, localised per market instead of rebuilt per market.
The corporate site couldn't change, so the page lives on its own domain. A qualifying form asks personal, business or government first, then language, format and province, and only then a name. Every lead reaches Salesforce with its answers attached. Now the template for all Berlitz Canada pages.
The business read booking drop-off as lead quality. Funnel analytics, 24 interviews in five markets and 8 moderated sessions showed motivated buyers abandoning mid-flow. Seven bottlenecks became a ranked redesign roadmap and the basis of a twelve-month product programme.
Six audience pages on one design system, each serving five markets from a single build. A three-question qualifier ranks leads before the booking calendar, and Smart Bidding trains on real bookings, not form fills.
Campaigns ran disconnected across 15+ countries with 85% of budget on local tactics. One system replaced them: a five-stage framework per persona, a brand belief spine, a 300-piece content matrix written for people, search and language models, and a three-tier measurement plan.
The first execution of the "Get Real" narrative: a global campaign theme that Berlitz Canada and the UK adopted as their lead story, with local campaign layers on one shared baseline.
Twelve social silos pulled into one hub-and-spoke system: one editorial calendar, a shared asset library and a monthly review feeding the next brief. In Canada, social was tied to the paid media calendar and became a measurable referral channel.
Creator collaborations, including the Sketches Between Languages series with Tilo, that introduced a 148-year-old brand to a new generation of learners.
An event identity built inside the Berlitz guidelines, plus the registration platform, so that 35+ assets came out of one system in six weeks. The identity was kept for the following year.
Led acquisition and retention for Liligo across SEO, conversion, app installs and paid. Built a server-side A/B testing product to measure every product change, and removed dark patterns that traded long-term engagement for short-term revenue, reducing CPL.
We built a KPI tracker app, then saw the client team lived in Excel. We retired the app, mirrored every formula into a workbook they own, and turned the engineering into a reusable dashboard product that flags any variance over 5%.
As co-founder alongside my Berlitz role, I advised on 0-to-1 product strategy, community growth loops and marketplace development. Version two launched too many features at once, so version three is rebuilt social-first, with paid services deferred.
Audits across five AI engines in buyers' own languages, then implemented fixes. A free full audit lost money on small deals, so diagnosis moved inside a paid sprint from €6,500 per market, keeping only the visibility score free.
Mojo runs on Claude with tool calling and retrieval. The model already knows coffee science, so retrieval draws on the user's gear, stash and brew history plus a catalogue synced from roasters. Onboarding was rebuilt twice to end on a taste profile.
Independent 72-hour verdicts on youth players from three analysts and a panel median. Selling trial places was dropped because it would compromise a neutral verdict. ScoutVerdict sells only the verdict.
Pet aftercare with vetted providers: cremation, pickup, memorials and grief support. Two interface modes, Reflective and Supportive, and a DACH-first launch in Berlin, Vienna and Zurich.
Owning growth and web for 70+ markets means deciding where one platform serves everyone and where a market needs its own answer, across B2C self-serve and B2B enterprise.
300+ experiments across enrolment funnels, a server-side testing product at eDreams ODIGEO, funnel research before rebuilds. I start from measurement: tracking before media, one KPI layer across markets.
I ship production AI products myself with Claude Code: retrieval, tool calling, data models and the guardrails around them. Engineers get a PM who can read the schema.
Three companies founded since 2014 taught me go-to-market, pricing and when to pivot. I bring that urgency to product teams without the chaos.