@Haodong have you heard about this before?
Yes. I think it’s one of the good startups in this field, along with neural concepts, emmi.ai…
I actually think semiconductor is a very good field for neural network based surrogate modeling. We have not touched much into this yet.
1 Like
They claim they can solve industrial semiconductor thermal analysis without customer data/fine tuning. I read their paper again (You can download from their webpage). It seems they are doing an multiscale thermal analysis of semiconductor package. I’m not exactly sure about the role of AI in their workflow. Maybe they are using AI as surrogates in homogenization process. Overall, it’s a multiscale analysis workflow. Maybe we can use MSG to predict the effective thermal conductivity. But I’m not familiar with chips design and analysis, not sure how to set up a model they are interested in. Maybe professor @Wenbin can bring in someone with more experience in this field and that would be very good for promoting MSG and swiftcomp.
@Jeffrey_Liu Maybe this is where AI surrogates can be very useful. I don’t know about semiconductors. But the success of Vinci4D might mean we can effectively sample the whole design space.
Have you downloaded the paper? you can upload it here. I went there and it ask me for information. @Haodong