The work, with the
numbers attached.

AI Mastery is Mindvalley's flagship AI program. Hundreds of students, hundreds of questions a week, and a program NPS sitting at −2. Most of those questions had the same thing underneath them: people felt behind and were nervous about saying it out loud.

Started as community expert, answering hundreds of student questions every week. That volume forced me to learn faster than any course would have. Then I got picked to teach live, and picked again. I rebuilt the hard parts as schematics: how RAG and vector stores hold memory, how Claude splits into chat, cowork and code, what Git and GitHub actually do. Breakout rooms every session, so nobody sat alone with a broken setup.

9.1/10 across 290+ post-session responses, 119 of them rated 9 or 10. Program NPS moved from −2 to 61. Nominated top speaker. One student wrote that it was the clearest description of RAG and vector memory she'd seen after several theory courses.

FEUP is one of Portugal's top engineering schools. Their faculty teach engineers for a living, and they wanted their own people fluent in generative AI before it turned up in every classroom and lab (which it did).

Built hands-on training for people who already think rigorously: prompting, then working with models on their own material, then agents. Everything ran live in the session on real work, so the tools got tested against the exact objections engineers raise.

An engineering school picked an outside trainer to teach its own engineers, which is the harder sell of the two. It also opened the door to the AI education work I'm doing now with universities and education hubs.

Canonical publishes Ubuntu. They hired me to run product for their AI/ML and MLOps line, which in practice meant Kubeflow: the open-source project Google's Vertex AI is built on. Two constituencies to serve, the Kubeflow community and Canonical's commercial strategy, and they don't always want the same thing.

Met Mark Shuttleworth, Ubuntu's founder, daily for a year. He coached me on serving an open-source community honestly while still serving the company funding the work. I led product, go-to-market and the OSS contribution side, wrote a large share of the documentation myself, presented Kubeflow live to 1,000 people, and ran webinars with Dell, Lenovo and NVIDIA.

I became the #1 contributor to Kubeflow documentation outside Google. It's also where the current mission came from: if writing docs could reach that many engineers, teaching could reach a lot more people than that.

Healthcare has the tightest constraints and the most to gain. Anchor wanted generative AI inside workflows that touch actual patients, working toward 2 million patient touchpoints.

I advise on strategy, tooling and adoption: which workflows are worth automating first, which ones to leave alone, and how to get clinical and admin staff to actually use what gets built. Adoption is usually the hard part, not the model.

Still in progress, so I'll be precise: 2 million touchpoints is the target we're building toward, not a result I get to claim yet.

After every talk, someone asked the same question: fine, but how do we put this into our workflows on Monday? A keynote can't answer that properly. So Andre and I built the arm that can.

Co-founded the consulting side, which goes into companies and implements AI where the work actually happens. Clients so far: FEUP, Lusiaves, SuperModular, Strato Automation and BIG Bank. Higher education, food production, industrial automation and banking.

5 organizations across 4 sectors, and a steady feed of real implementation problems that goes straight back into what I teach on stage. It's the reason my examples are current instead of recycled from a conference talk I saw.

People finish an AI course excited, go home, and hit the first error message alone. Two weeks later the habit is gone. I wanted the group itself to be the product.

Launched Ainauts Space as a private community, then designed the Ainaut Program around it: 6 weeks, hands-on, cohort-based, built so people ship something real and have witnesses when they do. Ran 4 weeks of promotion before cohort 1 opened.

13 students and roughly €9,000 in sales off 4 weeks of promotion. Cohort 1 stayed deliberately small, small enough that I could teach every person in it properly and watch what broke.

An AI summit audience has already heard every version of “AI will change everything.” I had the stage, 300+ people in the room, and wanted them leaving with something they could do on Monday.

Built the talk around rugged flexibility: a stable identity paired with the ability to adapt fast. Half field report from someone who's been in AI since the EPFL robotics labs in 2016, half practical map. I've since opened the European Innovation Academy accelerator with a version of it.

Nuno Couto messaged me afterwards to say it was the best talk of the summit that day, “com a dose certa de insights práticos” (with the right dose of practical insights). It's now my signature keynote, in English or Portuguese.