Ph.D. 2.0: Rethinking the Ph.D. Application

Note: this Ph.D. application season is over. I disabled the challenge problems below until next year. Thank you all who participated! The rest of this page is kept for informational purposes.

It's Ph.D. application season and as a faculty member on the admissions committee, I have been thinking about our admissions process and want to offer a way for applicants who appear unimpressive on paper can demonstrate their ability to succeed at a Ph.D.

An "Alternative" Application

Perhaps you want to pursue a Ph.D. with me as your advisor, but don't have a conventionally strong application because you have little or no prior research experience, or may not have three strong recommendation letters. Maybe you didn't publish as an undergraduate and develop relationships with three recommendation letter writers. Maybe you can't even write a very compelling research statement yet. So here I'll offer an alternative way for me to get to know you better.

First, I need to know you are a strong programmer and are ambitious. In my line of research, you need to build software to run experiments and prototype designs and analyze data in my research area. You can use either of the following as evidence:

  • A Github or other online repository of your source code demonstrating contributions to open source projects or impressive projects of your own.
  • An offer letter from Google or YCombinator (where you are the tech co-founder) which serves as evidence that you passed a challenging programming interview at Google, or that YCombinator believed you would be successful at developing your company product. (I'm not suggesting you interview with no intention of joining, but rather if you're considering multiple options.)

Second, you should have a research interest that overlaps with mine, and that you can start doing research independently. For this, I'm also offering two options:

  • A good software developer interview question is to ask the candidate to solve a real problem the company has, so I propose a challenge: show me a good start to one of these (download the zip):
    1. Gaze Prediction: Predict a person's eye gaze from their cursor movements. [hard]
    2. Crowdsourced Profs: Create a list of computer science faculty by crowdsourcing. [moderate]
    3. Tracked Geolocation: Visualize my minute-by-minute geolocation data over time. [easy]
  • The alternative is to demonstrate scholarship. Select 3 papers you aspire to write (ideally in my research area), summarize the key findings, explain why you chose these papers, point out any flaws you've identified, and discuss potential next steps of research.

Email me a single PDF file with the above, including your homepage URL and CV, to phdapplicant@jeffhuang.com and I promise your email will not go through my spam filters, it will automatically be marked as important in my inbox, and I will read it thoroughly. Do this before applying through the normal application process (deadline December 15, 2013) and I will respond within 72 hours with further instructions if I feel good about your chances, or that I don't feel you are ready (but you are still free to apply through the official process). I will also provide feedback if you wish; just indicate this in your email.


Disclaimer: The "Alternative Application" is an experiment, and I may decide to end it at any time. It supplements but does not replace the existing official Ph.D. application process at Brown University. If you already have a strong application, then by all means apply through the conventional process; I will review those applications anyways. The Alternative Application does not apply to any other faculty in our department; those applications will be evaluated using the conventional criteria. You still need a bachelor's degree. The opinions expressed here are my own and no other else's.

Background

The current Ph.D. application system has some problems on both sides of the table. First, the conventional application materials don't necessarily tell me whether you are able to perform the research, and mostly comprise statements written by you or your references rather than showcase your abilities. Second, as a new professor I need to reach out to students that I'd like to work with because students and their advisors in my field may not know me, and not even consider applying to Brown, a problem that larger universities like CMU, MIT, and Stanford might not have.

The Thiel Fellowship made waves when they offered undergraduate students $100,000 to drop out to build a startup. The first five years of a fully-funded Ph.D. at an Ivy League University costs roughly $300,000 to the advisor (this underestimates the total cost because a typical Ph.D. in Computer Science takes 6-7 years). So I'm making a bigger multi-year commitment both financially and with my time for you to "drop out" of joining a company, on very little information.

On the flipside, Ph.D. admissions are competitive; as a department, we admit about 40 students per 400 applicants. So an applicant may have a hard time putting together a strong application if they have meandered off the "optimal path", which is something like going to a top-5 undergraduate program, doing undergraduate research and internships with several faculty, being part of projects that end up published, and maybe even receiving the NSF graduate fellowship. These applicants develop a strong CV, 3 glowing recommendation letters, and professors clamor for them. Ph.D. computer science application advice suggests that students start research early and generate visible results, specifically, "I think that working in 3 different research groups is ideal" and "You should get started early, preferably on your first project at the beginning of your sophomore year."

But what if you are not this type of applicant; perhaps you didn't do undergraduate research and instead took interesting upper-division courses, or went to join a startup or a software development job after graduation. Then you've got an uphill battle; your recommendation letters (if you can even get three) will be unenthusiastic and you may get passed over for someone with research experience. It's perfectly rational from an admissions committee point of view---past success is the best predictor of future success.

I may be a case in point. My first round of Ph.D. applications were all rejected even with undergraduate research and several minor publications under my belt: 0/8. The next year I spent my time after work preparing a first-authored paper and luckily it was accepted for publication at a top-tier conference. I applied to 5 departments the following application season (hoping to not overburden my recommendation letter writers again), and was admitted into one school, where I ended up doing my Ph.D. For that I am grateful that I had a terrific experience there, but the rejections stung. In retrospect, it made sense: my recommendation letters may have been weak, my GRE scores average, my grades mediocre, and my research statement was not well coached.

So now being on the other side of the table, I want to put in extra effort to recruit strong students whose existing application materials might not be the best signals of future success.

Read my other essay on Adopting the Startup Culture for Research.


Back to Homepage