By New faculty member Misha Khodak started out studying math as an undergrad and switched to CS for his MS and PhD after a “throwaway comment” to a CS professor while looking for a math thesis topic. Khodak works on “developing the right modeling, training, and evaluation tools to make faster progress” in the natural sciences and engineering – and he’s excited to work on solving important problems with colleagues at UW-Madison.
Hometown:
Pennsylvania
Educational/professional background:
Previously I was a fellow at the Princeton Language Intelligence Institute. Before that I did my doctorate in CS from CMU.
How did you get into your field of research?
Ten years ago I was a math major in need of a thesis topic and emailed a CS professor about doing research in theoretical CS. At the end of my email, I made a throwaway comment that I’d done a course project on word embeddings. He was moving away from theory to machine learning at the time and happened to be studying word embeddings, so I ended up doing that for my thesis. It worked out well enough that I switched from math to CS for my masters and PhD.
What are your areas of focus?
Machine learning, specifically learning-augmented algorithms, transfer learning, and AI for scientific computing
What main issue do you address or problem do you seek to solve in your work?
I study how to integrate algorithmic and scientific data into machine learning models to make them more accurate, and how to integrate machine learning into algorithms to make them faster.
Please describe your work for people without a background in computer science:
We have seen the tremendous success of foundation modeling—training big neural networks on vast quantities of unlabeled examples—in developing AI tools for text, images, and audio. Such large amounts of data are also available in the natural sciences and engineering, and there are incredibly important problems to address, but we do not have a consistently successful way of harnessing the data. I work on developing the right modeling, training, and evaluation tools to make faster progress in these scientific domains.
What attracted you to UW-Madison?
The people in my department and others that I’m hoping to collaborate with were very welcoming and excited about solving important problems together.
What are you looking forward to doing or experiencing in Madison?
I’m hoping to go skating on the lakes in the winter.