Fellowship Applications

Applying to fellowships is quite a daunting task. When I was applying, it felt impossible to write about the work I would do as a PhD student when I — then an undergraduate — didn’t even know where I’d physically be in a year. The dearth of resources online and hodgepodge of sometimes conflicting advice I received from friends and mentors only worsened the problem.

I took the strategy of applying to nearly every fellowship under the sun and seeing what stuck. I don’t necessarily recommend that, but I also encourage those reading to apply to as many as they feel comfortable. Beyond the external funding, applying to fellowships really helped me clarify a coherent research question before starting my PhD. Although I ended up going in a different direction than what I wrote, it was a great exercise in writing coherently and cogently about a research problem.

I hope this post clarifies (a) the high-level strategies I tried to follow during fellowship applications, and (b) some more specific personal advice on individual fellowships.

What is a fellowship and why does it matter?

When I started looking at fellowships, it was unclear to me what a fellowship even was or whether they made any tangible difference in your graduate studies. I looked online and found intimidating resources like this GitHub page which had more fellowships listed than total faculty in some CS departments. Some of them I was eligible for and some not, but organizing the pile of opportunities was (quite literally) a tall task.

In short, fellowships provide external funding to cover your tuition and some or all of your graduate stipend.

For example, if you receive an NSF fellowship, NSF would pay your tuition (typically covered by your advisor) to MIT and a stipend which is typically below the standard rate for your department. Your department or advisor would pay the difference and (perhaps) a small bonus to reward you for winning the fellowship. So, if my departmental tuition were \$100,000 a year and my stipend \$50,000 then NSF would pay the full \$100,000 and \$35,000 to MIT. MIT would forward me this \$35,000 as well as, for instance, an extra \$20,000 to bring my total stipend to \$55,000, slightly above the standard rate. (It’s worth noting that these numbers vary by department and program, and not every department pays above the standard rate, even with a fellowship.)

External funding provides additional flexibility. For instance, I was fortunate to get a couple fellowships. Between them, I was able to navigate the graduate school selection process (post admission) with some peace of mind knowing that I wouldn’t have funding stress anywhere. This steered the conversations with potential advisors away from “do you have funding” and “how many students can you support next year” to “what topics are you thinking about nowadays” and “are there new research directions in your group?” To be clear, you should ask all of these questions regardless, but the fellowship can help you prioritize the more scientifically grounded concerns.

I’ve also heard of friends on fellowships who have found it easier to collaborate with researchers at other institutions (as your funding is not directly tied to your home institution). Lastly, the fellowships do provide a CV boost and potentially some networking opportunities.

What are some strategies I should keep in mind?

While I originally tried to write a one-size-fits-all essay, I quickly found that I had to scrap and rewrite portions of each essay to tailor it to the fellowship. The best resource I had were mentors — other PhD students and advisors — who themselves had gone through similar processes. Give them sufficient heads up and ask them to tear through your work. It was disheartening for me to know that some of the scientific ideas (e.g. “representation learning for biological sequences”) I had were too vague or imprecise, but the back and forth of editing helped me clarify the writing and my own ideas. I even caught up to the state-of-the-art in the field of computational genomics and AI for biology.

After my first round of messy edits, I started jotting down a few items for each fellowship:

1. Who will be reviewing my application? What level of technical depth is appropriate?
2. What personal qualities or traits are emphasized in this program? What words or phrases are highlighted on their website that flag specific qualities?
3. What types of people typically win this award? Is their work heavily interdisciplinary or field specific?

It’s great to have a rough template of ideas and essays common across fellowships, but it’s equally important to tailor your application. I found that answering the questions above provided an appropriate starting point for my applications. I knew that the DOE fellowship was focused on the high-performance computing aspects of my work, while for the NSF application I could emphasize the intrinsic value of the scientific questions raised by my proposal. I also let my mentors know how the fellowships differed in their focuses based on these questions.

The above three questions are the most universal advice I can give. Each of the fellowship-specific tips I provide below are based on the questions above.

NSF GRFP

Eligibility: US citizen or permanent resident, can apply once as an undergraduate and once as a graduate student

Timeframe: 3 years

Advice: Fortunately NSF is one of the few fellowships for which there are high-quality publicly available resources. The best resource is Alex Lang’s website, which contains over a decade’s worth of winning applications. NSF is probably the closest one to a grant proposal (or so I imagine), and I recommend providing the nitty gritty scientific details of your project. It’s also important to develop a clear, concise, yet well-scoped proposal that relates to your previous research. To this end, you should not have proposals as broad as “Develop ML Models Robust to Distributional Shifts” or as narrow as “Quantify Failure Modes of Diffusion Models Under ABC Rotational Transformations in Dataset XYZ.” Somewhere in the middle is the sweet spot, and the website can help you adjust.

NDSEG

Eligibility: US citizen or permanent resident, have at least three years remaining in your graduate students

Timeframe: 3 years

Advice: This one is a bit quirky in that they ask you to scan through their specific research goals (“Broad Agency Announcements”) and find one that suits your project. Make sure you choose wisely, because they certainly will assess your application at least partially based on the relevance of your work to this goal. Other than that, I’d generally emphasize how your research may enhance defense capabilities, even if you have to stretch your mind a bit. Do not just copy-paste your NSF essay and add the defense relevance at the end as I did; instead try to reframe your essay to highlight the relevance to defense throughout.

DOE CSGF

Eligibility: Undergraduate seniors and first-year PhD students who are US citizens or permanent residents

Timeframe: 4 years

Advice: This one is the most mysterious to me, but I can tell they care deeply about the coursework, so do not take that part of the application lightly! Again, try to reframe your essay to highlight its energy relevance — perhaps your work will enhance energy efficiency or provide new energy storage methods.

Hertz Foundation

Eligibility: Undergraduate seniors and first-year PhD students who are US citizens or permanent residents

Timeframe: 5 years

Advice: The Hertz Fellowship provides the most funding (5 years) and some networking opportunities. The Hertz Foundation spells out the values it cares about on its website: creativity, leadership, and innovation. Of the fellowships I applied for, I spent the most time contemplating how to frame my Hertz essays to highlight these traits. As someone who didn’t start finding interest in anything remotely biological until sophomore year of college, I made sure to emphasize how my somewhat unconventional journey into computational biology strengthened my work by providing a broad toolkit.

As for the interview, I found practicing with mentors and friends to be most helpful. I explained my work to those around me at different levels: in the technical weeds, high-level picture, and elevator pitch. I also ran a couple practice interviews with my advisor, who asked me to explain a number of technical concepts (e.g. explain a VAE, what is Hardy-Weinberg equilibrium, etc.). It’s acceptable and even encouraged to get flustered when you practice; in my opinion, the best preparation for the technical interview is to practice explaining thorny scientific concepts clearly.

Soros

Eligibility: Child of an immigrant or an immigrant

Timeframe: 2 years

Advice: I actually didn’t end up applying for Soros, mostly because I was swamped with other applications. That’s totally fine! While I did not apply, Soros awards fellowships to all types of graduate students: law, medicine, business, science, etc. As such, you should focus on broader themes of how your immigrant experiences shaped your work. I’d recommend focusing less on your specific research and instead trying to highlight your “immigrant” and “American” values that they mention on their website (e.g. hard work, innovation, creativity, cultural fusions, etc.).

Other fellowships

There are so many other fellowships, and I’d recommend applying to as many as you can. Some prominent ones include Knight-Hennessy (Stanford only), Quad (US, India, Japan, etc.), GEM, etc. There are also CS-specific ones sponsored by big tech companies, such as Google, IBM, Facebook, etc. The CS-specific ones often require departmental sponsorship and are only open to PhD students in the latter years of their programs. While I’m not so familiar with non-CS fellowships, I’m sure there are others you can find. Lastly, please remember that it’s totally fine to not be on a fellowship. MIT, for instance, covers tuition and provides a stipend to all students for their first year, regardless of advisor status, and this is common across many programs and universities. Afterwards, your advisor will cover funding (and feel free to clarify this with your advisor before joining). Good luck!