Product manager / systems thinker / builder
Accessibility, product systems, movement, and cross-cultural storytelling. At Microsoft I drive product direction for PowerPoint — the fundamentals millions of people rely on daily, and the Copilot features deciding where it goes next. On my own time I build across accessibility, AI, business, and the tools I want to exist.
01 — Day job
I hold both halves. The craft work that keeps the product excellent for everyone who opens it, and the Copilot direction deciding what it becomes. Neither one is allowed to starve the other.
Performance, reliability, editing craft, and accessibility for a product used by an enormous and enormously varied audience. This is unglamorous, compounding work — and it's what people actually notice. A feature nobody asked for cannot buy back a product that feels slow or fights you.
Driving where AI goes inside the product: what it should draft, rewrite, and restructure, how it grounds itself in the user's own material, and where it should stay out of the way. Scoping what the model can genuinely do well today rather than what demos well.
The interesting part of the job. AI investment competes directly with fundamentals for the same engineers and the same quarter. My work is making that trade-off deliberately and defensibly, instead of letting whatever is loudest win.
Great AI features and a great product for everyone are the same job. Treating them as competing agendas is how products lose both.
Read the full case02 — Selected work
Surfaces used by hundreds of millions of people, where a small design decision compounds fast.
03 — Approach
Four convictions that survived contact with large products and small ones.
Every product has a loud roadmap and a quiet one. The quiet one — speed, reliability, the thousand small edges — is what decides whether people keep the product open. I protect that budget, especially when something newer is competing for it.
"It feels better" is not a ship signal, and it's a particularly bad one for AI. I define what good looks like before the work starts, build evaluation sets from real user material, and hold the bar when the schedule argues.
I build screen-reader-first apps on my own time, and it changes how I spec everything else. If a feature only works when you can see it and aim precisely, it isn't finished — it's narrow. That test catches bad design long before an accessibility review does.
The senior version of this job is not shipping more. It's making sure a large group of capable people is pointed somewhere worth going, with a shared idea of what "done well" means — and then getting out of the way.
04 — Independent
Projects I run end to end, on my own time, across a deliberately wider spectrum than any one job covers: assistive technology, applied AI, community and business systems, and the tools I want to exist for my own growth. This is where I stay a builder rather than only a decider.
Shipping app · iOS
YouTube subscriptions turned into a channel-surfing experience for VoiceOver: videos narrated before they play, everything drivable from the lock screen or a Bluetooth remote. Designed, built, and taken through App Store submission solo.
Accessibility · iOS Open →Product concept
Route guidance for blind and low-vision travellers built around trust and pacing — supportive rather than a constant stream of instructions. An exercise in restraint as an interface principle.
Accessibility · Navigation Open →Product concept
Speech-led messaging: compose, review, and reply by voice, with far less interface to manage between you and the conversation. Where I think hardest about voice as a primary input, not an accessory.
Voice · Speech Open →Community & participation
Two systems about people rather than features: mutual support treated as infrastructure, and participation made intuitive and shared instead of resting on one exhausted organiser.
Community systems Open →05 — Beyond the work
Brazil, France, and the U.S. each left something in how I build. So did the things I do away from a screen.
Brazil
Origin, family, maracatu. Collective energy and the confidence to make something loud.
France
Engineering at Centrale, and a permanent bias toward restraint and detail.
United States
Scale, experimentation, and the habit of building in public.
06 — Contact
Then we should talk.