Getting to Know Pengtao Xie
Our team at Petuum is made up of incredibly talented people. This series will feature the engineers, managers, and creators that keep our company moving forward and make us proud of the work we do together.
Pengtao Xie is the project director of data service and solutions at Petuum, Inc. He is leading the product development and research in machine learning for healthcare. He is currently pursuing his Ph.D. in the Machine Learning Department at Carnegie Mellon University, working on latent space models and distributed ML for clinical decision-making. He is recognized as a Siebel Scholar and is a recipient of Goldman Sachs’ Global Leader Scholarship and National Scholarship of China. He received MS degrees from both Carnegie Mellon University and Tsinghua University, and BS from Sichuan University.
Why did you join Petuum?
Petuum aligns perfectly with my research and interests. I’m working on my Ph.D. in Machine Learning for healthcare, and a big chunk of our work at Petuum right now is building out Machine Learning models for healthcare in our beta product, PetuumMed. I’m passionate about putting my research into practice for people to use and, because of this lucky alignment, Petuum is a great platform for me to realize that dream.
Also, as a bonus, I’m close with the founders (Eric Xing is my Ph.D. advisor) and I get to work with a lot of my labmates and friends here!
What do you do at Petuum?
My team builds Machine Learning models and systems to solve critical applications in healthcare, such as predicting diagnoses and recession rates, image analysis, processing forms and highlighting important information, etc.
I also work on some of our other vertical solutions in areas I can’t discuss yet. I guess you could say that, in general, I work on applying ML and AI to certain industries and domains.
What’s the most exciting project you’ve worked on at Petuum?
I am personally most excited about our model’s ability to recommend treatments — when a patient comes in with their operation documents and admission notes and etc., our Machine Learning model takes all of those texts (even the ones written by hand!) as inputs and can predict what treatment course makes the most sense based on historical data. We have made a lot of great progress in our work in this area, and we wrote a paper and blog post about this specific capability called Predicting Discharge Medications at Admission Time Based on Deep Learning if people would like to learn more.
What’s been most challenging in your work so far?
The most challenging thing my team has faced is the need to collaborate closely with domain experts like physicians and hospitals to understand their workflow and requirements. We are trying to build something that healthcare professionals will actually want to use, so we need to work with them at every stage to make sure our models behave the right way and are helpful.
This has been a hurdle for me since I don’t have a background in medicine, and I need to learn and understand a lot of healthcare and medicine-specific knowledge in order to build the models well. I’m actually trying to teach myself about internal medicine — I’ve been reading a lot and talking with a coworker who trained as a medical doctor.
Do you have any advice for people interested in pursuing similar work?
My advice for people interested in researching healthcare and AI is to learn from my challenges — in order to do the work well in this domain, it’s very important to know both Machine Learning and medicine. Spend time learning both, and you will have a much better perspective than pure Machine Learning researchers and be able to point your project in the right direction.
What do you love to do outside of work?
Pittsburgh has an amazing symphony orchestra! I go there often. Most recently, I saw two violin showpieces, Ravel’s Tzigane and Chausson’s Poeme.
I enjoy being active and doing sports like running, skiing, and badminton. And I also love to read — I’m currently reading a few different books, but my favorite right now is The Great Gatsby. It is a little difficult for me to read it in English, but I enjoy the writing style and story.
We’re hiring! We’re growing quickly and looking for highly trained and talented technical staff, including engineers, architects, and more. Contact us at firstname.lastname@example.org.