Why are you involved in the NUM?

As part of the Medical Informatics Initiative, we have been working for seven years to create the technical conditions in Germany to enable the exchange of data between university hospitals. It therefore made sense to utilise the platforms created there in the NUM to launch clinically relevant projects and thus bring our research into broad clinical application.

Where do you see the greatest opportunities if all university hospitals conduct joint research?

I see many opportunities here. One of them is that we typically only recognise patterns in data when we have a large amount of patient data at our disposal. But we can only collect this amount of data if it can be collected at all university medical centres through networking and collaboration in the NUM.

And this is not only relevant for the diagnosis of diseases, but also for further developing therapies and guidelines and generally improving our understanding of disease mechanisms. For me as a data scientist, Artificial Intelligence in particular offers special opportunities that make it possible to take preventive measures and improve diagnoses and therapies by making predictions from data.

 

Tell us a technical term from your job that sounds exciting and that only the real experts understand! What does the term mean?

A technical term would be "large language models" (LLM). ChatGPT, for example, would be such a large language model. Most of us use such models regularly these days, but few of us know what is really behind them.

They are generative AI models that are based on a huge amount of data, use a large number of language documents and then reproduce them. These models are generative because they generate something new from something known.

 

What excites you about your job?

I am fascinated by the variety and the numerous challenges that the job of a data scientist entails on a daily basis.

A lot has happened in the field of data in the last 20 years. There have been drastic developments on the IT side. In terms of hardware, we can now calculate huge models with limited resources. On the software side, two or three years ago, experts could not have predicted what is possible today, namely being able to communicate with large language models at eye level. As a human being, this gives me pause for thought, while as an IT scientist it really excites and fascinates me.