Gandal Foglia

Designed and shipped an n8n + Python + LLM automation system for a client, eliminating about 120 hours a year of manual data processing, a CHF 6,000 annual saving verified in the contract. The project ran July to September 2025.

I continue to advise informally on process discovery, AI opportunities, and digital transformation for SMEs.

1. Most SMEs sit on enormous manual work they've normalized. Once you look, the wins are obvious. I now start every engagement with a process audit.
2. Digitization before automation. You can't automate a paper process. The foundation comes first; automation is the second step, not the pitch.
3. LLMs in production are unpredictable. 97% stable still beats a human. I build with fallbacks and ship. Perfect is the enemy of deployed.

Five consecutive summer internships at one of Europe's leading research institutes in AI and machine learning. Data annotation and collection across face recognition, vein biometrics, presentation attack detection, and audio analysis.

The first internship started when I was 15.

1. Machine learning is mostly data and statistics. The glamour is downstream. Data quality is the first constraint in any technical project.
2. Rigor, seen from the inside. Patience, precision, high standards. Working alongside top researchers set a bar that carried into everything after.

Co-founded and organized two TEDxYouth events in Martigny: 300+ attendees, 10 speakers, a team of four volunteers. I was 16 during the first edition. Master of Ceremonies in 2019; stage direction and full program curation in 2021. The 2020 edition was cancelled by COVID.

Speaker selection and coaching, day-of logistics, marketing, licensing with TEDxMartigny/Neurhone, school partnerships.

1. Incomplete information, unmovable deadline. You learn to structure what you know and map what you don't. I now write a known-unknowns list before any kickoff.
2. Different work styles need explicit roles. I define scope and ownership at the start of any team project. Ambiguity is a cost, not a feature.
3. Going on stage doesn't get easier. But you stop fighting the adrenaline.

A digital loyalty app for small businesses, replacing paper punch cards for local cafés and restaurants. Built and shipped to the App Store and Google Play independently.

The name is redacted because of a trademark claim filed after release.

1. Digitization doesn't solve fundamental problems. Loyalty has to exist first. I now validate the underlying behavior before designing anything around it.
2. Validate before building. A few coffee chats with potential customers would have revealed the real issue before any code. Customer discovery now comes first.
3. Register your trademark immediately. Someone filed after seeing the app. Never again.

A multilingual Secret Santa web app: wishlists, budgets, automatic matching, email notifications. Built to solve the annual coordination problem and to learn new tech.

1. Tech isn't always the solution. Some coordination problems are social, not logistical. I now ask whether a problem actually needs software.
2. Friction kills adoption. Requiring accounts for group actions was a fatal flaw. I design account-free flows first by default.

A minimalist iOS memory journal. Three-second videos or dual-camera photos. No feed, no followers. Offline-first, with iCloud sync.

1. Native performance matters. iOS frameworks gave real advantages for camera and video. The right tool for the constraint, not the familiar one.
2. Technical debt compounds fast. Unaddressed optimization problems snowballed. I now fix performance issues at first encounter.
3. Scope should match resources. Video processing needed more than I had. I now size projects to what I can actually finish.