Machine Learning PhD Graduate 2018 Highlight of the Week – David (Wei) Dai- Sr. Director of Engineering at Petuum
Third part of #MLCMUGrads18, where we are celebrating our PhD Graduates for 2018 where each one of our Machine Learning PhD graduates will be on the spotlight, due to their contributions and achievements in Machine Learning.
Follow along through social media #MLCMUGrads18 or by following our RSS feed.
David recently graduated with a PhD in Machine Learning after successfully defending his thesis on: “Principled System Designs for Large-scale Machine Learning” – David’s research focuses on large scale machine learning that lies at the intersection of both, systems and theory. He emphasizes that most “Big Data” researchers on machine learning have only focused on one or the other. David currently works as the Senior Director of Engineering at Petuum, where he is building a scalable machine learning (ML) platform with his team, designed for enterprises to easily create and manage complex machine learning workflows and models.
THE CONVERSATION
How do you think that Carnegie Mellon has prepared you for the future?
Computer science at CMU is very interdisciplinary, and I got to collaborate with experts from different fields like machine learning and distributed systems and think across a diverse range of abstractions, from hardware, distributed systems, programming languages, statistics, to applied math. These experiences really help me to pick up new subjects more easily, such as healthcare AI, product designs, and blockchain.
What is your favorite quote?
“Most people overestimate what they can do in one year and underestimate what they can do in ten years.” - Bill Gates
What is your main takeaway from your trajectory with Carnegie Mellon University?
For me a big realization at Carnegie Mellon University is that a PhD is actually not that different from doing a startup. I needed to find an advisor who’s like an investor and is aligned with my research interest. I had to figure out the right research problem to address, and solving the wrong problems leads to lower research impact or impasse. (Similarly, most startups fail because they address the wrong problems). I also needed to find and convince people with the right expertise to “join my team” on the projects, and apply problem solving along the way. Above all, I had to work really hard and really believed in what I’m doing. All these entrepreneurial trainings helped me to reinvent myself in a rapidly changing technology landscape.
What are you planning to do now that you have graduated?
I am currently leading the engineering team at Petuum, which is a Pittsburgh-based machine learning startup founded by my advisor Professor Eric Xing. Throughout my PhD I’ve been working on large-scale machine learning system and algorithms, which is a great match to what Petuum is doing on bringing cutting edge AI to enterprises. The company has recently grown from <10 people in 2016 to now 100+ and we are running strong, so it’s very exciting!
Do you have any advice for current PhD students in our department?
Picking research questions usually determines 50-80% of the research outcomes, and the rest is how to solve them. This may sound like an obvious statement, but it wasn’t obvious for me for the longest time. So pick carefully! Try to vet with people from different perspectives (not just your advisor) your research problems, and not just about how you plan to solve it. And remember to apply proper exploration vs exploitation in research and in life (restaurant choices).
David is an inspiration to our current students and prospect students at Carnegie Mellon University. We couldn’t be more proud of David for successfully graduating from the Machine Learning Department with a PhD in Machine Learning, his achievements and contributions to AI, Machine Learning and CS will just continue to rise with Petuum, and future endeavors. Remember to stay tuned to #MLCMUGrads18 in social media or follow our RSS feed to stay updated with new highlights as they come along.
