Portfolio

AskPancho.ai · my own product

AI for vetting second-hand flats.

The problem: Buying a home is the most expensive decision most people ever make, and they make it once or twice in a lifetime. Nobody teaches them how to read a listing or what to look for at a viewing.
What we did: The app reads the listing, studies the area and guides the buyer through the viewing: it tells them what to ask and requests photos of anything that looks suspicious. I designed it, coded it and launched it myself.
The result: A fully working web app, on the market within a few weeks of the idea. It has kept improving ever since, based on what users do.

AiScope · NGO

The problem: Many rural areas have a microscope, but nobody who can read the sample. Diagnosis takes days, or never happens.
What we did: An Android app that connects the phone to the microscope, guides the health worker through capturing the sample and stores it with labels for training the AI. Designed for people who have barely used a smartphone. I coordinated a team of more than 40 professionals from different countries: developers, designers and doctors.
The result: The largest known public dataset for training malaria diagnosis models. Awards and talks at international conferences.

Fit 4 Life · sports event timing · Switzerland

The problem: A company that times sporting events had data on thousands of runners and wanted to offer them more than their race time.
What we did: I sat down with the head of the company, interviewed athletes, hired the team and led the build of the first prototype: an AI coach that turns each athlete's goal into a plan and keeps adjusting it based on the training they actually do.
The result: A working app that helps athletes prepare for a specific event.

exocad and Align

The problem: A technician models each restoration by hand from a scan of the patient's mouth. It's slow and expensive, and the result depends on who does it.
What we did: I built and led a team of more than fifteen people: developers, dentists and data labellers. We created our own tool for labelling 3D scans and trained an AI that generates the restoration ready to print or mill.
The result: exocad's AI made it onto Align's roadmap. The rest is covered by an NDA.

Oklusion · dental clinics and labs

The problem: Orders went back and forth by message and phone call. Details got lost, there were mistakes and delays, and the patient was left waiting.
What we did: I interviewed dentists and dental technicians, defined the product and led the design and development of a web app where the order, the quote and the conversation all live in one place. We were a small team and we moved very fast.
The result: Everything was up and running in six weeks, at a very reasonable cost.

Grupo Alvic

The problem: Spain's third-largest office furniture manufacturer sold only through distributors and wanted to reach end customers without setting up a separate company.
What we did: I built the online sales department from scratch: web shop, marketing, logistics and an in-house customer service team, coordinated with the factory.
The result: Mueblesdeoficina.com selling online, with its own team and processes that kept working without me.

Internal tool for a software team

The problem: Translators, reviewers and developers passed files around by hand. Every release of the application shipped with untranslated or outdated text.
What we did: A platform with roles and an approval workflow, synced with GitHub, so that whatever is approved flows automatically into the version customers use. I built it on my own.
The result: Within two days it was working and the chaos was over. By day eight it was deployed, after four rounds of changes with the users themselves.

More examples of applied AI, with demos you can try, in the cases section.