Sci-Phy (“Sci-Fi”) Lab will be recruiting its first cohort of PhD students during the upcoming 2025-2026 application cycle. If you are interested in working with us, this guide will help you better understand our research philosophy, interests, and what we’re looking for in applicants.
We work across the full robotics stack. So we’re particularly drawn towards people that enjoy operating at the fringes of their intellectual comfort zones, and expanding them over time. Our main goal is to bring our research to life on real-world physical systems whenever possible.
Please apply through graduate admissions and indicate me as a faculty member you’d like to work with. I am affiliated with the CS department, and I will prioritize admitting students that have applied to the CS PhD programme this cycle. Additionally, I will begin reviewing applications by first considering candidates who have listed me as their first-choice advisor.
Note: Due to the large volume of emails I receive about PhD admissions, I’m usually unable to reply to most of them, though I do read many. The guidance below is to reduce that email debt and offer applicants a window into my approach to research and thinking. If you still feel the need to email me (and do not mind not hearing back), use the word labubu somewhere in the email subject so it makes past my filters.
I often encourage my mentees to think critically about some of these questions before deciding whether, and where, to join a PhD program.
Advising philosophy
The best thing a PhD program affords you is the time and space to deeply think about problems you care about, and to solve some (or all) of them. As an advisor, my goal is for mentees to develop research independence and create meaningful impact in the communities they care about. I find myself most satisfied when we make meaningful progress towards longstanding problems in robot learning, as opposed to churning out publications in high volumes. I am particularly interested in problems that we (academics) are uniquely equipped to address.
While robotics is my central focus, my work often draws from fields like computer vision, graphics, and cognitive science. I encourage my students to read broadly and think critically, even if that sometimes means spending more time ideating before running the next experiment.
I believe that writing is a form of structured thinking. My mentees will likely spend significant time writing (idea summaries, planning documents, etc.) before they write code.
This advising philosophy will undoubtedly change across students and career progressions.
Would you consider yourself a hands-on or hands-off advisor?
I don’t think anyone can be entirely one or the other. My level of involvement (“hands-on”-ness) depends on several factors:
Indeed, for most projects, I might be more involved in certain phases or in particular components of the project (this is especially true if this is a collaborative project involving other faculty / researchers).
As an example: if you’re an early grad student (years 1–2), we will work closely on brainstorming topics, diving into implementation details, and will be co-writing manuscripts. Over the next few iterations, you will increasingly find yourself driving parts of this process more, until you are eventually able to drive the full research cycle yourself.
Would I need to have published research to apply?
Not necessarily. Publications certainly help, especially when they are in areas of interest to me. However, I value (and will actively look out for) other demonstrations of creativity and initiative. Examples include:
What might be some research directions this first cohort of students could work on?
While I always have a ton of project directions in mind, my advising style is not to impose a project or a direction on a student, but rather to come up with something we’re both equally excited about. I like to work on what my students like to work on.
In the near term (next year to two), I am excited about some of the newer waves in robot learning research (e.g., generalist models for manipulation and navigation), particularly from the perspective of understanding what these models know (and don’t), and to leverage our prior knowledge of the world (3D, physics, etc.) to make them far more capable.
I am also increasingly interested in multisensory perception, having dabbled in tactile and audio sensing during my postdoc.
For representative projects in the public domain, please look at Krishna’s featured publications.
What will you look for in an applicant?
I am particularly interested in candidates who demonstrate creativity, initiative, and a clear sense of research direction. In your statement of purpose, I encourage you to dedicate roughly 30-50% of the content to discussing the research problems you hope to pursue. I’d like to understand why they matter, the unsolved challenges therein, and how you might begin to approach them. This section is far more informative to me than general statements about your interests in robotics or the desire to pursue a PhD.
Beyond machine learning experience, I value applicants who bring in breadth across (or outside) the robotics stack. For instance, if you have experience and prior work in areas such as classical control, robot design and manufacturing, 3D perception, game design, fabrication, etc., please be sure to mention this in your application. Additionally, if you are able, enclose pointers to project materials (reports, code, etc.) that I can look at. Depth in any of these areas, coupled with curiosity and openness to interdisciplinary thinking, is a strong signal of fit for the lab.
What materials would I need to prepare to apply?
Please refer to the Whiting School of Engineering admissions website for more details. As of October 2025, there is a USD 75 application fee (which I understand can be waived in some cases). The GRE is not required to apply to the PhD program.
What could be some caveats of joining Sci-Phy Lab?
The main one is that I’m not particularly interested in writing papers that merely add incremental value to existing results. This higher bar can mean publishing fewer (but potentially more thoughtful!) research artifacts (papers, blog articles, code, etc.).
Another caveat is that my intellectual curiousity often draws me to combine insights from multiple (and sometimes disparate domains), and this style may not appeal to everyone, and that’s perfectly okay.