Agenda
🕡 6:30 pm // Start of Event, Welcome
🕡 6:45 pm // 🎙 Talk 1: Hey Siri, when will you become truly useful?, Sherin Grimbergen, Cloudflight
Voice recognition and speech generation have improved dramatically in the last few years; lower latency, more robust recognition (of dialects), increasingly natural voices, etc. These advances come from specific architectural shifts. In this talk, we’ll look at the current state of the art: what technologies are driving these improvements, why they work, and where their performance breaks down in real world scenarios. Based on hands-on experience, we’ll discuss the technical bottlenecks that still limit applications. Other than customer support agents, there are still very few voice applications consistently delivering good user experience. The goal is a grounded assessment: how far are we from effective voice applications, and what would need to change for Siri and her successors to get there?
🕡 7:30 pm // 🎙 Talk 2: Transformer–xLSTM Ensembles: Achieving 2nd Place in the 2025 PhysioNet Challenge for ECG-Based Chagas Disease Detection, Angus Nicolson, Medical University of Innsbruck
Diagnosing Chagas Disease — a condition that can lead to heart failure and arrhythmias — is crucial, and automated diagnosis from non-invasive, low-cost ECGs is a promising solution. However, there are two major machine learning challenges to overcome: a severe class imbalance (2% positive cases) and a high-level of label uncertainty from self-reported patient diagnoses. In the 2025 PhysioNet Challenge, our team from the Digital Cardiology Lab at the Medical University of Innsbruck achieved 2nd place through large-scale pretraining of multiple foundation models. We utilised the complementary strengths of Transformer models (leveraging attention mechanisms) and xLSTMs (a recurrent architecture well-suited for time-series data like ECGs). In this talk I will summarise the challenge, the problems we encountered and how we overcame them, along with some technical details of our solution.
🕡 08:15 pm // Food, Drinks & Networking
🕡 09:15 pm // end
Location
Cloudflight, Bürgerstraße 1-3, 2. OG
6020 Innsbruck
Date & Time: 25.11.2025, 6:30 – 9 pm
Entrance: free – register HERE
For more information contact michal.janecek@cloudflight.io
✨ Everyone’s welcome – get ready for an evening of innovation, great people, and inspiring discussions!
Let’s shape the future of AI in Innsbruck together! 🤖💡