Like the image above, we have known issues and big opportunities for improvement in our current model (DOOG-E V1). We have released V1 in Beta because want to get early feedback from users as we build V2 of our Large Medical Model DOOG-E for any use cases or other problems that we have not yet uncovered.
Our first publicly accessible model, DOOG-E V1, is a toy model relative to the next version which will be much larger. Currently V1 is only trained on a sub population of the United States of about 1M patients. V2 will have well over an order of magnitude more patients, and in totality will be approximately 1000x larger. The increase in size is due to more patients, longer histories, and more data types. We are rigorously testing V2 to more accurately present under which use cases you can expect production level performance.
If you’re interested in seeing how Transformers built on medical event data can be helpful to your organization or care and are willing to test and provide us feedback, please send us a note at hello@genhealth.ai requesting access to our V1 model. We make no warranties or claims of accuracy for our V1 model so this Beta is purely for research and development purposes.