Trustworthiness has taken center stage in today’s AI discourse. This AWS-sponsored meetup invites participants to explore what trustworthy AI means and how it can be realized in medical applications.
📅 Event Schedule
Friday, March 13
Doors open: 6:00 PM
Presentations begin: 6:30 PM
⸻
✨ Spotlight Talks
-
Trustworthy AI for Healthcare: From Benchmark Performance to Clinical Reliability
🎤 Sara Steiner
AI Researcher, Know Center
-
Translating Prediction into Performance: ML-Guided Co-Processed Formulations for Inhalation Treatment of Tuberculosis
🎤 Sarah Zellnitz-Neugebauer, PhD
Senior Scientist, Research Center Pharmaceutical Engineering (RCPE)
⸻
📍 Location
UNICORN Startup & Innovation Hub
Conference Deck
Schubertstraße 6b
8010 Graz
👉 For better planning, please sign up in advance.
⸻
🤓 Abstract : Trustworthy AI for Healthcare: From Benchmark Performance to Clinical Reliability 👩⚕️
(by Sara Steiner)
Artificial intelligence is increasingly moving beyond research labs and into clinical settings. Among these developments, large language models are being explored for clinical documentation, patient communication, and medical decision support. While these systems demonstrate strong performance in controlled settings, real-world healthcare is far more complex.
In medicine, accuracy alone is not enough. AI systems must be reliable, safe, understandable, and aligned with the needs of both clinicians and patients. Trust cannot be assumed - it must designed.
This talk explores what it takes to build trustworthy AI in healthcare. Drawing on ongoing projects at the Know Center, it highlights how model development, system architecture, and empirical validation — together with close collaboration with various stakeholder groups — shape whether AI improves care or complicates it. The future of medical AI will depend not only on smarter models, but on building systems that clinicians and patients can trust.
⸻
🤓 Abstract: Translating Prediction into Performance: ML-Guided Co-Processed Formulations for Inhalation Treatment of Tuberculosis 💊
(by Sarah Zellnitz-Neugebauer, PhD)
Treating tuberculosis (TB) is a long process that requires taking multiple antibiotics daily over a long period of time. Typically, antibiotics are administered oral, what often leads to side effects and antibiotic resistance.
Here, administering the antibiotics via inhalation directly to the lung, where the disease is predominantly manifested, could help. However, often just blending the antibiotics together does not work efficiently and often drugs show poor solubility. Therefore, we propose to administer two antibiotics in form of a co- amorphous system (COAMS). COAMS are stable single phase systems comprising of two compounds that stabilize each other so that the system overall shows increased solubility, bioavailability and stability compared to just the single APIs.
However, finding matching drug pairs for COAMS is not trivial. To simplify this, we are using AI to find matching API pairs and design "all-in-one" inhalers that deliver several drugs in a single, fast-acting powder. Since testing every drug combination in a lab is slow and expensive, the team uses a machine learning model to predict which drug combinations will be the most stable and effective. To ensure the AI's trustworthiness, its predictions were rigorously validated against real-world lab data, achieving a high accuracy rate of 79%. In fact, the model successfully identified two specific TB drug pairings that were then proven to work in physical laboratory tests. By combining the speed of digital predictions with the reliability of experimental confirmation, this approach creates a "gold standard" for drug development. This verified AI-driven process saves significant time and money while ensuring new treatments are safe and effective.
Ultimately, this builds a bridge between digital innovation and proven medical science for better patient care.
⸻
‼️‼️‼️‼️‼️‼️
We will provide finger food and would like to avoid unnecessary food waste.
If you are unable to attend after signing up, we kindly ask you to let us know in advance or cancel your registration.
Please note that photos will be taken during the event and may be shared on our social media channels. If you prefer not to appear in any photos, please let us know - we will, of course, respect your wishes.
We are very much looking forward to welcoming many interested participants and engaging in a lively discussion with you.
Let’s continue bringing all minds together! 🤗
Your WAI Vision Team Styria,
Claudia Kumpitsch, Suchita Kulkarni, Isabella Danda, Jelena Saric
