Applying to grad school is easy; picking which programs to apply to is hard.

TL; DR: Pick an advisor, not a school

Disclaimer: I am no expert in admissions. This series of blog posts is a hastily organized braindump resulting from abundant introspection, and from discussions I’ve had over the last few years with some very bright prospective grad students. Much of this is my view of the process and holds (if at all) only for North American PhD programs in ML/CV/Robotics.

This article is for you if you really want to do a PhD and are wondering how to pick programs to apply to. It is also for you if you have received multiple admit offers (congratulations!) and are deciding where to go.

Special mention to Aditya Kusupati and Vikram Voleti for proofreading an earlier version of this article.

“So, what schools do you plan to apply to?” is a question I frequently ask prospective grad students. Invariably, the answer is “top-10 or top-15 schools according to XYZ’s ranking scheme”. I’m not a big fan of this line of thinking – over the years I’ve come to realize that there’s deeper factors influencing your grad student life than merely the ranking of your prospective school.

I enlist many such factors in this post; I believe prospective grad students might not otherwise consider these when applying. Each factor has its own pros and cons: pick what works best for you.

Here’s an excellent blog post by Tim Dettmers touching on related themes which prompted me to pen down my thoughts.

And, another blog post by Danish and Mansi that prompted me to finish this article, having shelved it midway.

Choose an advisor, not a school

Personally, the only recommendation I have—and I cannot stress this enough—is to choose an advisor, not a school. Your advisor plays a pivotal role in your grad school experience and eventually, your career. A good advisor won’t just help you with your research; they will help you broaden your professional network and help you stay focused, healthy, and happy.

Spoiler alert: There is no one-size-fits-all when it comes to advisors. Every grad student has a different style of carrying out research, and thrives best when the advisor best matches (or complements) them.

Most students look out for the obvious when scouting for advisors: a research area fit. While that is a necessary condition, it is far from being sufficient. An advisor’s personal traits, work style, and bandwidth are often more crucial aspects to consider.

  1. Hands-on vs hands-off: A hands-on advisor will closely work with you in formulating ideas, figuring out most of the details, and perhaps even help implement them. Often they will write bulk of the manuscript, tidy up the figures, and help draft rebuttals if needed. This might be extremely helpful (and essential) if you prefer being mentored at close quarters. A hands-off advisor will expect you to come up with ideas, implement them, write a manuscript up and will often proofread and provide broad advice if restructuring or additional results are needed. While not a good fit for students that prefer working up close, hands-off advisors encourage you to bloom into independent researchers, setting you up to carve out your own research agenda.
  2. Early-career vs post-tenure: Usually, early-career researchers (i.e., postdocs and assistant professors) tend to be more hands-on compared to researchers post-tenure. Early-career researchers often need to pour in a lot of time and resources to scaffold their group, while post-tenure researchers often have everything in-place (and perhaps support staff to look after administrivia). Post-tenure researchers are well-connected and frequently get invited to talk about work; this might fetch your work visibility. On the flipside, this means their time is stretched thin – you will often not see them for weeks on end. From an admissions standpoint early-career researchers tend to have more openings, which might increase your likelihood of being hired by their group. Post-tenure researchers may/may not hire grad students each cycle depending on various factors (e.g. going on a sabbatical / other long-term vacations).
  3. Adjunct or mixed academic-industry roles: Several advisors are riding the AI wave, often splitting time across industry and academic labs or taking on full-time industry roles while maintaining adjunct professor status with the university. Such advisors often have great connections with industry – this could be an easy ticket for you into your advisor’s organization; could also foster collaborative work. Then again, such advisors usually are pressed for time working two jobs. Moreover, collaborating with industrial organizations might impose a few restrictions on what you work on and how you may share it.
  4. Other preferences: In addition to the above, some advisors may impose additional restrictions such as showing up at the office during pre-specified work hours, the number of holidays you are allowed per year, and whether or not you could go away on summer internships.
  5. Co-advising: Many programmes offer you to the flexibility to be co-advised. This might allow you to experience the best of both worlds (e.g. be mentored by early-career and post-tenure researchers) and expose you to multiple groups with a diversity of ideas. Note, however, that you will still need to choose your primary advisor, and that it might sometimes be difficult to agree on a direction that would maximally excite all your advisors. In general, I wouldn’t worry about (not) being co-advised; especially if both your advisors are at the same school or if they often collaborate. In the long run, it may not matter whether or not you have all of them formally listed on your thesis.
  6. Locked-in advisor vs rotation: While some programs have your advisor locked in when you join the program, others expect you to find a suitable advisor within a year of your admission. Both approaches have their merits. You might realize after joining a program that you and your advisor are a mismatch – hence it is important to ensure there are alternate advisors available and that they are equally eager to work with you. Hearsay: Programs that do not explicitly match advisors during admissions ensure that at least 2-3 prospective advisors express interest in a candidate before extending admit offers. Even in such cases, advisors often find that students’ interests tend to evolve over the course of their PhD.

Above all—as rightly said by Jorge Rodriguez– choose an advisor who is kind. Kindness makes up for any deficiencies in the above. Advisors who are kind will keep supporting you no matter what.

Spend time researching the group, your labmates

Beyond your advisor, your research group and labmates have big impacts on your grad life. Good research groups are inclusive, foster ideas, and offer the opportunity to get meaningful feedback on your work. Spend time researching the structure of the group and analyzing whether you would fancy spending 3-6 years there. I’d suggest looking for:

  1. Composition: Get a sense of the size of the group (this can vary from a handful to several hundred!). In particular, what are the various levels of researchers in the group (postdocs, senior PhDs, early PhDs, masters, undergrads, interns)? Does the group also have non-technical staff (e.g. office/lab managers to help manage workspace and admins to help with filing reimbursement claims, etc.)? Does the group actively hire part-time visitors/interns to help support research projects? Is the group diverse enough?
  2. Collaborative nature: Does the group frequently collaborate with other groups at the university? With groups at other universities? If so, are there any frequent collaborators (particularly ones you might fancy working with in future)? A good way to sense this is by sorting through their recent research. Equally important is to get a sense of how the intra-group collaborative structure is (does each student have their own project? do multiple students collaborate on an initiative?). Working at a school that tends to often compete can be toxic to your PhD – you could end up spending more time in authorship tussles or idealogical debates as opposed to work.
  3. Networking/Outreach: You might also want to see if the group offers networking and outreach opportunities beyond publishing and presenting research. For instance, does the group actively organize events (conferences/workshops)? Does the group have a frequent set of visitors from academia/industry to give talks and network?
  4. Student support: Does the group/school run initiatives to keep grad students happy? Are there support systems to identify and deal with mental-health issues (e.g. mental-health first-aid)? Are there internal groups that discuss more than just resarch (diversity and inclusion problems, buddy programs for newly admitted students, other fun student activities)?

Nature of funding; expected research directions

Most schools guarantee funding to PhD students at the time of admission. Your advisor(s) are often responsible for having obtaining the requisite funding before extending you an offer.

  1. Source of funding: The source of funding may play a role in determining the nature of your research. If you are being funded by a specific project with fixed deliverables, it is possible that your research agenda/role is locked in at the time of your admission. However many project proposals are open-ended to allow for the student to scope their own research problems. At the other end of the spectrum are labs with no-strings-attached funding, which offers you complete control over what you might want to work on. This is often the case when you are funded by graduate fellowships. It is hard to get a sense of your source of funding before you join a university, but expected research directions are a must-discuss aspect when you visit or talk to prospective advisors.
  2. Extent of funding: Some positions might only offer partial funding with the understanding that you will earn the rest of your funding via TAships or other admin work. Furthermore, some of these PhD positions might also have fixed durations after which your funding will expire. Make sure you know the extent to which your studies are funded.
  3. Resources: Does the school have an abundance of resources to support you? E.g. if you’re doing ML, you probably want to go to a school/group that has an abundance of GPU compute. This will ensure that resource constraints won’t let you compromise on your research – not being able to finish up a submission because you didn’t get a GPU job allocation can be frustrating. Likewise, if you aspire to work on robots, go to a lab with lots of them.
  4. Perks: It might also be important to know if there are any perks to your funding; for instance my fantastic advisor sponsored an all-expenses-paid trip to a top-notch conference during the first year of my PhD just for me to get a feel for the venue and the research landscape.

Quality of life

And then there are other quality of life choices such as the ranking of the school, activities that the school and/or the geographic location offers you, and the impact on personal life.

  1. Brand/ranking: I have come to believe—over the years—that the ranking of the school does not significantly impact your eventual career trajectory. As rightly pointed out by Danish and Mansi in this blog post, “you are the driver of your PhD”. All the above advisor/group factors applied, your work speaks for you! In some cases though, the brand and/or ranking of the school/group might help you. Particularly in communities where credentials are overvalued (such as when founding companies), being in a location that offers you geographical and/or networking advantages serves as a launchpad.
  2. Geographic factors: You might also want to consider geographic factors such as the distance from your home country/city, and the overall likeability of the place.
  3. n-body problems: Occasionally, prospective students might need to solve n-body problems (i.e., figure things out with their partner/immediate family) – an important personal life aspect often underlooked. E.g. if you’re in a long-distance relationship, or you have close family that needs attending to, you might want to choose a school/advisor that offers you the flexibility to work remotely as often as needed.
  4. Stress: A lot of the above factors when added up with the weight of your graduate studies can impose a lot of stress. Avoid all avoidable stress. Pick schools/groups/advisors that will help you maintain a healthy work-life balance.

A word of caution

It’s important to note that grad school life is more than just publishing research; you often find yourself immersed in the game of publish-or-perish and your overall ambitions/goals become muddled. Pick a program that allows you to do what you’d really enjoy doing all the while keeping you physically and mentally healty – and you’ll reap rewards in the long run. I hope this article has poked you to think about a few aspects of your prospective grad school life that you haven’t already. Drop me a line (Twitter DMs/mentions welcome) if you have thoughts to share.

References

  1. Tim Dettmers’ excellent blogpost titled “How to Pick Your Grad School”
  2. This Tweet by Zachary Lipton
  3. Danish and Mansi’s blog post titled “University Rankings for PhD Applications” (this motivated me to work on my unfinished article)

Here are a few other articles in this series.

  1. Writing a statement of purpose
  2. Demystifying letters of recommendation