Four positions I hold about doing this work, each one paid for by the project underneath it. Who I am and where I have worked is on thefront page.
I shipped a dashboard redesign that lifted engagement 40% and still cannot tell you whether it solved the problem, because nobody captured the baseline. Task time, click depth and error rate take an afternoon to instrument.
On my own product I wrote the prohibitions before the features: never promise someone they will hear a voice again, never put a user count on a page, never say unlimited about a thing with a ceiling. Honest copy is nearly free to specify and expensive to retrofit.
At Volkswagen I argued for an experience that cost the business money. Fleets could hide their data from the dealer network completely. Designing that refusal properly is what made the default trustworthy rather than just tolerated.
My AI feature scored 14 out of 14, and two of those answers contained things that never happened. A mechanical check cannot see a fabrication that is correctly formatted. Somebody has to read the output.
Not on the front page, because each of the four case studies there shows something these would only repeat.
AI crypto research platform · 2023 to 2024
An AI SaaS that pulled fragmented crypto data into something an investor could act on. Launched to 2,400+ beta users, then closed. Inference cost more than the market would pay for it. I keep it here because a product that died of its own unit economics taught me more than most of the ones that lived.
NFT marketplace for surf · 2022, freelance
A marketplace built so surf fans could own moments without knowing what a wallet is. Onboarding to first purchase in under a minute, for an audience with no blockchain literacy and no patience for acquiring any. 2,000+ beta users.
NASA Space Apps, mobile award with IBM · 2015
Interacting with astronauts in real time through augmented reality, over a feed of planets, probes and live mission events. I ran the ideation, the wireframes and the interface, and defined the AR experience.
Medical device interface · 2016 to 2017
A clinical risk tool. The design problem was presenting a probability to the person it describes, which is the same problem I met again a decade later and handled better the second time.
One honest sentence about what you actually do, not the category it belongs to.
One honest sentence about what you actually do, not the category it belongs to.
One honest sentence about what you actually do, not the category it belongs to.
BA, Multimedia Design
Universidade da Beira Interior, Portugal
2011 to 2014
Portuguese and French native, English fluent.