Foundation model for electronic medical records

Electronic health records (EHRs) are easy to access more than ever. They are a gread data source for various studies, a gold mine of information. However, due to its tabular form it is not trival to apply them directly to modern machine learning algorithms.

Although it can be time-consuming, Human doctors can extract necessary information needed to understand a patient using EHR. One can think of a model that can do the same, to extract features of a patient from his medical records in tabular form.

I aim to do exactly that, to develop a model that can learn a representation of each patient given large volume of EHR, a foundation model for EHR. Building a model that can correcly understand data is the necessary basis for many future research.