Harvard Statistics Interviews Lucas Janson about Tenure Promotion

The Harvard Department of Statistics recently announced on October 27, 2025, the promotion of Lucas Janson to tenured professor.  Widely recognized for his influential research in the areas of high-dimensional statistics and machine learning, Prof. Janson is also highly valued for his excellent undergraduate and PhD mentorship and dedicated service to our department and the broader statistical community.  We interviewed Prof. Janson to learn more about his passion for and highlights in research and teaching (please see the edited and excerpted interview below).

Career Path:

1. What is the significance of this tenure promotion for you and your career path?

Janson: The immediate significance for me was knowing about all the support from my current and former students, and my colleagues at Harvard and beyond.  Career-wise, this promotion means that I can continue doing what I love for the rest of my life.

2. At what point in your education did you decide that statistics and academia were the right path for you?

Janson: In college at Stanford, I was interested in math and physics and initially planned on completing a master's in financial mathematics.  However, I changed course and completed a master’s in statistics because I felt that a statistics master’s would be more broadly applicable. My statistical coursework and research during the end of my undergrad career confirmed this choice. 

After graduating, I worked in finance and quickly realized that it wasn’t for me and decided to apply to PhD programs.  During my PhD program at Stanford, I ultimately determined that I wanted to stay in academia because I found the work to be a lot more fun and rewarding. As a professor, I appreciate the independence I have in selecting my projects and collaborators and feel that the environment is the right fit for me.

Research and Collaboration Highlights:

3. Your work spans high-dimensional inference to contributions in reinforcement learning and robotics.  Describe how your work has evolved over time, including some of your highlights. 

Janson: The driving force behind a lot of my work since my PhD has been in high-dimensional inference. An exciting aspect of this area is that it allows researchers to use statistical methods to take advantage of powerful, computational, non-parametric machine learning methods. It's a modern research area both because there are all these new, little-understood predictive algorithms to use statistical inference on, and because, for the most part, my work does not rely on specific knowledge about complicated algorithms.  In the future, people can use much more complicated algorithms, but my flexible wrapper methods that go around the algorithms would still work to provide statistical guarantees. 

My interest in robotics and reinforcement learning started when I was a first-year PhD student and found the opportunity to work with a roboticist, Marco Pavone, who works on algorithms that help robots plan how to move.  The question of how you design an algorithm that allows a robot to move through a cluttered environment ignited my interest in reinforcement learning. More specifically, I became interested in learning in an environment that contains unknown factors because this becomes a statistical problem.  Reinforcement learning is really online learning based on what is in the environment as the algorithms take steps and adaptively collect data by using the information learned.  Right now, I'm interested in uncertainty quantification for reinforcement learned data because it’s an interesting challenge; it does not have any of the nice statistical properties that most statistical methods rely on.

For future research directions, I plan to continue in two research directions, broadly high-dimensional inference and uncertainty quantification for adaptively collected data. Interpretability for artificial intelligence (AI) is an additional area that would be exciting to explore.  AI can be treated as a complex model that produces data, and new statistical tools will be important for understanding these important models. In both my existing lines of work, I develop new statistical tools to allow users to understand complex data, so AI would be a new domain to work on this problem. 

4. Your work is also very collaborative. For example, you have collaborated with colleagues in scientific fields including genetics, political science, and climatology. Describe a recent collaboration that you found to be rewarding.

Janson: A collaboration that I would like to share from the past five years is one with microbiome researchers.  I started working with Curtis Huttenhower in the School of Public Health, and then continued with one of his former postdocs, Siyuan Ma, who is now an assistant professor at Vanderbilt.  Two papers have come out of the collaboration; Siyuan is on both papers, and the second one is led by one of my current students, Ritwik Bhaduri.  At first, I thought the problem was just another version of variable selection, but I learned so much from my collaborators about all the statistical implications of the compositionality of the data and the sparsity of the data.  As a result, the collaboration opened up completely new statistical questions in variable selection for me that were really fun to answer.

Teaching and Mentoring:

5. The contributions of your pedagogical innovation are evident based on the courses that you’ve thought of and designed: Stat 195 Introduction to Supervised Learning, Stat 184 Introduction to Reinforcement Learning, and Statistics 305 Statistical Consulting.  What motivated you to create these courses and what is their value?  What other curricular ideas are you interested in pursuing?

Janson: For the courses Stat 195 and Stat 184, I built these courses based on the observation that there was a need for machine learning courses in the department, particularly at the undergrad level.  Stat 184 was co-created with Sham Kakade, which was an awesome experience because he knows a lot more about the field.  Now these courses are central to our new Machine Learning Track for concentrators, which is great to see.

The idea for Stat 305 Statistical Consulting came from my participation in statistical consulting as a PhD student when I learned about how to communicate with non-statisticians and the broader world of statistical practice.  When I came to Harvard, there was interest in the course from the students, and the faculty were supportive, which led me to create the course.  Stat 305 has become integrated into the department and into the PhD qualifying exam process, which has been satisfying.

In the future, I could imagine teaching a course in randomization testing.  There are some very elegant arguments in this area that would make it accessible to a wide range of students without a math-heavy background.  It's a pretty big area, touching on everything from really classical stuff, like Fisher's Exact Test, to modern hot topics like conformal inference.

6. Many of your undergraduate and graduate student mentees have received departmental or Hoopes prizes for their theses. You have also exposed students to research, who otherwise may not have pursued it as a career path.  How would you describe your approach to mentoring students and what have you learned in the process?

Janson: My approach is to talk with students about the types of problems that excite them, instead of just assigning them a problem.  Once we are working together, I try to get to know the students to learn the best way to shepherd their career, and hopefully, a budding interest and excitement for statistics research moving forward.  This can be challenging; I’ve learned from different parts of my career that the type of support I needed at certain stages was quite different.  All of my mentees are so unique, and I try to figure out how to be the best mentor for them, which, hopefully, has been good enough!

Harvard Community:

7. Since starting in the Harvard Statistics Department, what have you valued the most about the Department and Harvard in general?  

Janson: The people at Harvard have made the experience so wonderful.  When I was a very junior faculty, I had a lot of support from senior colleagues, who helped me figure out all the different dimensions of being a professor.  It has also been fun getting to spend a lot of time with junior colleagues with whom I feel a great sense of camaraderie.  Of course, the students are amazing; I’ve had the opportunity to work with some really stellar undergrads and PhD students in my group. 

Additionally, the staff and faculty design a lot of great events that I love to participate in, such as Data Adventure Day [a day of statistics and data science activities for ~100 public high school students] and the Stat 300 PhD Student Seminar. Stat 300 has been a great way to get to know the students and faculty and the work they are doing—it’s something I wish I had as a grad student!  In general, it’s nice that the department is a little bit smaller, which makes it feel like a close-knit community.