Production ML, automation, research, product. A few of them I built purely for the fun of it. Open any card for the full story, numbers and all.
My main project: a system that merged our marketplace data with public sources and told each salesperson when and where to act, delivered entirely over email: briefings out, commands back in the replies. It found money in the noise the reps had no time to mine.
An ML tool that scored sales opportunities and suggested the margin to quote for the odds a rep wanted, tuned to the rep and the client. I validated it live against the sellers' own predictions, and it won executive backing for AI.
I met the head of quality while networking and told him I could turn six years of messy NPS comments into something useful, fast. Thousands of free-text answers, a dozen languages, shop-floor spelling. I built the pipeline that sorted them into praise and criticism, by theme and by country.
I built this coffee shop's online store from scratch, the family sold it in 2024, and in 2026 the new owners hired me back to automate it with AI. Pricing and restocking now run on their own; support and marketing are getting there.
My father had the coffee idea, and the customers weren't coming. I moved the shop to Shopify, took over the marketing, and grew it past CHF 500K in its first year.
It started as a networking chat with a sales director. His team was screenshotting live ads by hand, hundreds a day, to prove to advertising partners that the campaigns they'd booked were live. I built the tool that captured that proof and checked it against the plan.
A finance manager who looked after the marketplace sales numbers kept receiving spreadsheets that crammed several separate tables onto one page. I couldn't change how the files arrived, so I wrote a parser that finds each table and lifts it out on its own.
A 2022 rebrand of the family coffee shop: from a faceless, template-built storefront to an identity with a point of view and one promise, love the customer. Brand thinking picked up at P&G, put to work on a small business.
I co-authored an IEEE paper (SCIS&ISIS 2018) on the Probabilistic Orienteering Problem: routing under uncertainty, where a stop might be gone by the time you reach it. We trained a neural network to predict how much Monte Carlo sampling a candidate route needs.
Published at EAAI-18, this classroom activity turns a room into a live dataset: students photograph their own hands, an Android app and web service pipe the shots to a server, and a convnet learns them in front of everyone, overfitting and recovery and all. I built the app, the web service and the model-serving glue.
A ROS and Qt tool that replays a drone's real flight against its planned path, live, so you can see and measure how far it strays. Built for a 2017 study putting an ultra-wideband positioning system through real flight trials, where GPS is hopeless between buildings. Rebuilt here as an in-browser demo.
A supervised classifier for these surface defects already existed and worked. My MSc thesis asked a harder question: could an unsupervised model catch the same defects without ever being shown a labelled one? Autoencoders learn what a good surface looks like and flag whatever they can't reproduce.
A 3D, explorable map of what I've studied: hundreds of concept bubbles from my own Coursera transcripts, wired by typed links and grown by a local LLM whose every edit I can preview and undo. It grows most days, and I use it daily.
Local-first, it pulls Swiss job postings, scores each one against my own profile, and rolls the gaps into the two answers I care about: which skills to study next, and which roles are worth pursuing. A supervised browser extension handles the applications.
A founder came to me with a buzzword-heavy idea, blockchain and a decentralised back-end, for helping people in need through local shops. I cut it down to a working MVP and built it, acting as product owner and lead developer for a small team.
A game with a player-driven free market, an arbitrage seam, and no API to read it. So I built a pixel-by-pixel price reader, a bot fleet to feed it, and a nightly engine that told me what to buy and sell. At its peak I was sitting on about three billion in in-game currency.
A non-profit in Ticino gathers real, difficult experiences and wanted them to stay with strangers. I built the pipeline that turns each account into a short story and an image, under one rule the model couldn't break: make it readable, invent nothing. The pairs went online and onto exhibition walls.
An automated first pass for a local-media observatory: it scrapes a region's papers, scores each one with sentiment and bias heuristics plus an LLM summary, and hands the volunteer team a reviewable shortlist. The editorial call stays human.