Machine Learning & Artificial Intelligence Meetup Bern

Agenda

• 17:45 Open Door at ZID Bernapark, Bernapark 28, 3066 Stettlen (RBS Station Deisswil, approx. 13min from main station Bern)
• 18:00 Welcome
• 18:05 Talk 1: Building AI application using Spring AI and spec-driven development grounded on company knowledge by Altin Hani & Filipe Inacio from Break GmbH
• 18:30 Talk 2: Do Patents Kill Papers? A large-scale ML pipeline says They Don't by Emma Scharfmann from Artifact
• 19:00 Apéro with Pizza and drinks
• 19:45 Closing

About Talk 1: Building AI application using Spring AI and spec-driven development grounded on company knowledge
Spec Driven Development is more relevant than ever. In this session, we’ll show how well this pattern pairs with Spring AI to build a custom chat that is grounded in your company’s own knowledge. Instead of producing generic answers, the chat can reliably draw from internal content and provide results you can trace back to the source.
We’ll outline a pragmatic workflow—from a clear spec, to connecting knowledge sources, to implementing a simple Spring Boot setup. Along the way, we’ll cover the key building blocks (ingesting content, preparing it, and reusing it for retrieval) and what to keep in mind to turn a quick demo into a useful internal assistant.
By the end, you’ll have a concrete blueprint for using Spec Driven Development and Spring AI to build a company-knowledge chat in your own environment.

Speaker: Altin Hani & Filipe Inacio from Break GmbH

About Talk 2: Do Patents Kill Papers? A large-scale ML pipeline says They Don't
When a scientist patents technology, does it detract their focus from science work? Answering this question remains difficult owing to a lack of comprehensive data on individuals who both publish and patent.
We constructed and made public a new dataset of "Pasteur's quadrant researchers" (PQRs) who both publish science and patent technology at some point in their careers. We compared two large-scale databases—scientific papers and patents, each containing tens of millions of records—using vector embeddings derived from titles and abstracts combined with structured metadata fields. A supervised machine learning classifier was trained to identify documents written by the same individual, with validation against Google Scholar profiles.
The dataset shows that PQR papers and patents correlate with greater novelty and future citation, relative to the work of those who only publish or patent.