CAIML #32

CAIML #32 happened on September 3, 2024, at neuland.ai AG.

Agenda

18h30 Open Doors

19h00 Welcome & Intro

19h15 Islam Torki - Artificial Intelligence Engineer at neuland AI: Lessons from the Frontline: Sharing Insights and Lessons Learned from Implementing Generative AI in Real-World Applications, Including Firsthand Experiences, Key Successes, and Practical Solutions

Join me as I share our experiences from Neuland AI, from the hurdles we’ve faced to the breakthroughs we’ve made, and what we’ve learned along the way. We’ll dive into how well RAG is working for us, pointing out where it shines and where it can trip up - like when it comes to parsing, retrieving, and synthesizing information. These areas can be tricky, and we’ll talk about how we navigate these challenges. Evaluating RAG pipelines is another topic we’ll cover. Sometimes it’s straightforward, but other times it gets complicated and requires more creative approaches. If you’re interested in the nuts and bolts of using Gen AI in practice, this session will offer valuable insights and practical takeaways.

19h50 Peter Lenz - Professor of Theoretical Physics at Philipps University Marburg - Decoding Complexity: Unveiling Gene Patterns with SDCM

In the complex world of biology and medicine, understanding gene interactions is a daunting task. Scientists use advanced technology to take vast measurements, but deciphering meaningful patterns is like finding a needle in a haystack. Our new Machine Learning method, called Signal Dissection by Correlation Maximization (SDCM), tackles this challenge head-on. SDCM sifts through massive data to identify unique ‘signatures’ of characteristic gene expressions within subsets of samples. What sets SDCM apart is its ability to precisely handle complex, mixed and multi-layered signals. We applied SDCM to study diffuse large B-cell lymphoma, a type of cancer, and uncovered new gene interaction patterns directly linked to patient outcomes. These patterns proved to be more accurate predictors of survival than those identified by other methods. Beyond cancer analysis, SDCM has shown its versatility by interpreting field trials of soil improvement products in agriculture. By separating real signals from noise, SDCM provides a robust solution for noisy, complex datasets.

20h20 Networking with food and drinks provided by neuland.ai AG