Apply, apply apply! Even if you don’t feel like you’re ready, the more you apply the more likely you are to get a fellowship and the more practice you’ll have for fellowship and grant writing in the future. Here, I discuss my process and thoughts about applying to the NSF GRFP and DOE CSGF in the 2025/2026 cycle.

Intro

During undergrad and in the first year of my PhD, I’ve applied to a lot of research fellowships. Applying to these awards has been one of the most important steps I’ve taken to improve my research, ability to formulate projects, and communication skills.

Much like writing code forces you to understand every step of the program you’re implementing, writing a fellowship application forces you to understand every component of a research project. You have to justify and explain why it matters, how it fits into a larger scientific landscape, and communicate it to people outside your subfield. I often understand a project best after I’ve had to write about it.

In this post, I’ll mostly be compiling resources for two national fellowships that I was fortunate enough to have received: NSF GRFP and DOE CSGF. There are many more awards out there that I haven’t applied to or was rejected from and as a result, don’t feel qualified to speak on, but hopefully these resources and anecdotes are generally helpful. Selected components of my application materials for these awards are linked as a resource as well!

National Science Foundation Graduate Research Fellowship Program (NSF GRFP)

The NSF GRFP is a graduate fellowship that funds you and your ideas for three years of graduate school if you are doing a PhD in STEM. The application is split into two main parts: the research statement and personal statement. Both of these statements must include sections addressing two broad criteria: intellectual merit and broader impacts.

There is so much I could say about this application - but I believe that others have covered it very well (some of these other resources are linked below in the Examples + Guides section for each fellowship). But there are a few pieces of advice I got from others and learned myself while applying that I’ll be walking through here.

Research Statement

One exercise I found surprisingly fun was drafting each fellowship proposal around a project I could envision pursuing in whichever rotation lab I was working in at the time. While this was more difficult than directly branching off of my undergraduate research, it was a great exercise in learning what the writing process looks like in each lab and coming up with research questions (as “practice” for writing a thesis). I also highly encourage (if you are in a rotation-based program) asking your rotation PI to write you a letter of recommendation.

Before writing the proposal, I found it super helpful to review the funding priorities document for the National Science Foundation for the year (publicly available on their website). I used these priorities to frame my project, and emphasized the application of my project’s findings to supporting these funding priorities in the broader impacts section of my research statement.

If the project you’re interested in doesn’t directly speak to the annual funding priorities, it may be helpful to chat with PIs or other graduate students at your university to discuss what future implications of the project you’re proposing, or work a few steps down the line look like. Chances are, even if it’s not immediately apparent, your work likely relates to these priorities.

Others mention this, but there are many MANY formatting guidelines. Be sure to follow them all or your application may be returned without review. Be very very careful about how you frame projects involving human health or disease. NSF generally supports fundamental scientific research, while disease-focused or clinical research is typically funded through NIH. Many projects involving human or disease-related research are still appropriate for NSF, but the framing should emphasize advancing fundamental knowledge rather than developing treatments or clinical interventions.

Coming Up With Novel Research Questions

I want to emphasize that coming up with a novel research question before starting or at the very beginning of a PhD is hard. I struggled with it a lot, completely writing out two alternate projects before landing on the one I submitted for the NSF GRFP. And even then, the project was heavily revised from my original draft. I also procrastinated quite a bit on this application. Starting early is definitely the best way to avoid the many late nights I spent working on it.

I had some prior experience developing research questions in undergrad, but coming up with one in a completely new field wasn’t easy. For both the NSF GRFP and DOE CSGF applications, I started by identifying assumptions, methodologies, or common practices in the fields I was rotating in that bothered me. Maybe a silly way to come up with a research question, but it worked for me. If a particular assumption, methodology, or interpretation consistently bothers you, chances are it bothers others as well.

For my NSF GRFP application, I focused on the issue of group averaging in human fMRI research in response to naturalistic (movie-watching) stimuli. For my DOE CSGF proposal, I focused on how the relative scarcity of clinical overnight EEG data limits our understanding of the diversity of electrical activity patterns associated with sleep and insomnia.

After identifying these concerns throughout my rotations, I conducted a focused literature review to see how others had approached these issues in the past and where gaps remained. For both applications, I also had computational techniques I was excited to apply - for NSF GRFP: topological data analysis and DOE CSGF: data simulation, dynamical systems modeling, and predictive machine learning. The resulting proposals were ultimately built around using these computational approaches to address the limitations I had identified in human neuroimaging and sleep psychiatry research.

I am aware this sounds like a pretty straightforward approach - it wasn’t in practice. I went back and forth between ideas countless times before settling on the projects I ultimately proposed. Nonetheless, this is roughly the process I followed, albeit with plenty of dead ends, revisions, and feedback from members of my rotation labs along the way.

Personal Statement

In your NSF GRFP personal statements, you are tasked with explaining your personal story and motivation for pursuing a PhD, justifying your choice of undergrad institution, explaining your past research experiences and relevant awards you’ve received, and how you as a scientist contribute and plan to contribute to broader scientific and lay communities. It’s a lot to explain in three pages. As you might notice, I reused a lot of sentences and paragraphs from prior essays I’d written, especially from my statement of purpose for graduate school.

As you’ll also see with examples online, people take advantage of the relative flexibility of formatting for this statement (as long as you have bolded Intellectual Merit and Broader Impacts you’re mostly good). I decided to lead with a “hook” of my personal story and motivation for studying neuroscience before jumping into what I want to do with my PhD, then covered my prior and current research experiences in the Intellectual Merit section. In a lot of examples you’ll find online, people namedrop specific research programs they attended, bold their significant research responsibilities and accomplishments, and self cite their publications and presentations. I did the same and found it helpful. Bolding and italicizing important points make it easier for reviewers to quickly get through your application.

Similar to PhD applications, in explaining each of my research projects, I focused on answering:

1. How did I get involved in this project and how does it contribute to my broader research vision?
2. What question was I trying to answer?
3. What was my specific contribution?
4. What did we find?
5. What did I learn? How did it inform my next steps?
6. What came out of the work (papers, presentations, awards, etc.)?

I also gave substantial context for why I did the research I did and what I took from each experience. I also used more space to explain the projects that were most relevant to my proposed research direction.

In general, I approached my personal statement as a persuasive essay. In the introduction, I aimed to demonstrate that I had a personal motivation for studying the problems I proposed to work on. In the Intellectual Merit section, I explained how my past experiences prepared me to pursue those questions. Finally, in the Broader Impacts section, I demonstrated that I had the skills and experience needed to ensure that my research and other activities as a scientist would have meaningful impact beyond academia. I discussed my teaching experience, awards and feedback I had received for teaching, and initiatives I led to encourage broader participation in STEM. I then included a paragraph that tied these themes together:

“My primary goal as an undergraduate was to bridge the fields of physics, mathematics and computer science applied to multiscale problems in psychology and neuroscience to build a multidisciplinary background ideal for studying computational neuroscience in my PhD. I have built this background through my coursework and extracurriculars, but more importantly, the research projects I actively sought out and led in each discipline. Moving forward, I aim to integrate these disciplines to develop theoretical frameworks and models that better explain how complex cognitive processes emerge from brain dynamics and impact our everyday lives.”

Then, I described how I aim to ensure my research has “broader impacts” in my PhD (through teaching and mentoring roles and my “extracurricular research”) and why getting the NSF will enable me to carry out my broader research vision (in my PhD and beyond).

Final NSF GRFP Thoughts

If you’re applying to graduate school this year or are currently a first year, I highly HIGHLY encourage applying. It’s a great funding opportunity, a great way to get an award on your CV (if you get Honorable Mention or are awarded the fellowship), and in the case that you don’t receive any accolades, a great exercise in coming up with a novel research question and an approach to tackle it.

Other Examples + Guides

• Alex Lang’s Website: An incredible resource with countless examples, guides to writing the statements, etc.
• Ian Liu’s Repository: Has a lot of compiled NSF GRFP writing resources
• Examples from the GEMs Program: Also a ton of resources!
• Stanford Grantwriting Academy Repository: My personal favorite resource (though examples are only accessible if you’re currently at Stanford)
• Yaniv Brandvain’s Repository: More sample essays!

Department of Energy Computational Science Graduate Fellowship (DOE CSGF)

The DOE CSGF (Science + Engineering Track) is a unique fellowship opportunity as it blends a substantial graduate stipend with full tuition and health insurance support, a 12 week “practicum” (internship) at a DOE National Lab, and the requirement/opportunity to take advanced coursework in mathematics, computing, and your specific scientific/engineering field.

Throughout the application, it is important to emphasize your interdisciplinary computational ability and interest. You want to convince the reviewers that you don’t just happen to do science that sometimes involves computing. But rather, you are a computational scientist who NEEDS computing to solve the scientific problem you’re working on. This is important for two reasons: this fellowship funds researchers that aim to advance computational science technology (Math and Computer Science Track) and applications (Science and Engineering Track) and you will have the opportunity to use some of the nation’s most powerful supercomputers for your research available at National Labs (in both your practicum and throughout your PhD). Emphasizing your interest in using these resources and their potential to advance research in your topic of interest is thus important to a competitive application.

I think prior experience with HPC systems is extremely helpful, both for building a competitive application and for demonstrating that you understand the capabilities and limitations of large-scale computing resources. Similar to the NSF GRFP, I also found it helpful to look at Department of Energy funding priorities from the year I was applying. Documents stating these priorities were publicly available for the DOE as well.

I do wish that I had looked into practicum sites more before applying. While national lab websites are difficult to navigate, the fellowship keeps a list of current and alumni projects and practicum sites since the program’s inception in 1991. Abstracts and recordings of talks from annual fellow presentations at the DOE CSGF program review are also available online in fellow profiles. This is a great resource for learning about what students accomplish with the fellowship. The program also does a great job at tracking outcomes for its fellows, available in the profiles of graduated fellows on the program website.

In addition to being a unique graduate fellowship, the DOE CSGF has a unique application. In the application, you’ll be asked to propose a “program of study”, write a personal statement, and essays reflecting on your past experience with HPC and proposing your (HPC-based) research plans in graduate school. In the next sections, I will discuss these application components in more detail.

Program of Study

In the Program of Study section, you’ll be writing about the courses you plan to take the first three years of your PhD that fulfill the application requirements. I found it really helpful to talk to computationally leaning neuroscience PIs, postdocs, and graduate students at Stanford to discuss what courses would be most critical for me to take. I honestly wish I’d spoken to even more people about courses before applying.

That being said, the program of study doesn’t need to be perfect when you submit your application. I was accepted to the fellowship with a request for revisions to my program of study. What does matter is that it is tailored to your specific project. For example, my original proposal was based on using mathematical modeling to study heterogeneity in insomnia presentations, so I proposed courses relevant to signal processing (for handling the EEG data), specific modeling methods I wanted to incorporate (dynamical systems and predictive machine learning), and domain specific courses (relating to sleep and utilizing diverse recording methodologies in cognitive neuroscience).

However, it is important to note that your math/stats classes must come from a math/stats department and computer science courses from computer science departments. These considerations are where my Program of Study revision requests came from. You are also free to modify your Program of Study over the course of your fellowship as long as you complete the requirements before the start of the fourth year of your PhD.

Statements

There were three short essays required in this application: “Use of Computational Science in your research”, “Personal Statement”, and “Research Statement”. My application materials for this fellowship are linked above. My research statement is omitted, but a high level summary is included in the “Coming Up With Novel Research Questions” section and below.

Use of Computational Science Statement

The Use of Computational Science essay had two parts: (A) describing the most complex computational calculation or problem you’ve tackled and (B) describing what you could accomplish if given computational resources 100 times more powerful than what you currently have access to.

In part A, I think it’s critical to describe why HPC or other computational resources were necessary and/or useful for solving your research problem. Like the discussion of research experiences in the personal statement of the NSF GRFP application, I think it’s important to briefly describe the context of your research problem, what you contributed to working on it, and outcomes of the research question and why they are important. In part B, I think the justification is more important than framing the problem you could solve with 100x more powerful computational resources. This question assesses your awareness of the capabilities and limitations of HPC systems.

Personal Statement

The Personal Statement is very open ended for this fellowship. Unlike the NSF GRFP personal statement, I treated this essay as an opportunity to discuss who I am outside of research and how that shapes the way I approach science. I discussed my personal motivation for pursuing a research career and how returning to my art hobby helped me recover from burnout near the end of college, emphasizing how integral creativity is to my scientific process. I’ve seen a lot of variety in samples online. In my opinion, the idea is to demonstrate qualities that make you a good researcher in different areas of your life and to demonstrate that you are a multifaceted person as well as researcher.

Research Statement

Finally, in the Research Statement, I think the goal is to demonstrate that you can construct a research question and an approach to studying it that requires HPC resources. The question also comes in two parts: (A) describing an important, outstanding scientific or engineering challenge in your field that requires computational science and (B) the particular problem that you wish to pursue in your research and its impacts on science, engineering, and society if addressed. The rest of the application, including discussion of your past research experiences and disseminations, extracurriculars, awards, and prior coursework justify your qualifications for completing the project. I chose to define a broad problem in computational psychiatry: the lack of treatment direction and specificity resulting from a categorical diagnostic system. And my proposed project was development of an alternate approach to studying psychiatry through the lens of sleep data.

Final DOE CSGF Thoughts

Evidently, I spent a lot more time thinking about writing the application than I did thinking about what it would actually be like to be a DOE CSGF fellow. In retrospect, I wish I had spent more time exploring practicum opportunities, national labs, and the experiences of past fellows. While the application is important, you’re ultimately applying to join a community of computational scientists. If your research genuinely depends on advanced computational methods and you are excited by the prospect of working at a national lab (both in your PhD or potentially after), I really recommend applying.

Other resources!

• Kyle Felker’s Repository: Some great examples!
• Louis Jenkins’ Application: Example full application
• Mala Radhakrishnan’s Application: An interesting perspective on the application process

Looking back…

I think the greatest value of these applications and the others I completed this year wasn’t the funding itself. Writing them forced me to think deeply about what problems I want to work on, why they matter to me and society at large, and how I plan to approach them and whether that makes sense. Even if you don’t receive the fellowship, going through the process alone is an invaluable experience as a computational scientist. So if you’re eligible: I say apply! And as always, happy to chat if you have any questions or want to speak about these fellowship applications further.

General Resources

• Olga Botvinnik’s Perspective: A really great read!