I recently submitted a NSF CAREER proposal — my first attempt [1]. In the process of putting my materials together, I was fortunate to be able to learn from successful proposals written by friends and colleagues [2]. I was also fortunate to get feedback from mentors [3]. Trends emerged, and I thought I’d distill and share here on the off-chance that someone might find it helpful [4].

Disclaimer: I submitted to the Security and Trustworthy Cyberspace program, and requested examples from others who I knew submitted to that program. I expect much of what I learned applies to other programs within the CISE directorate’s purview, but your mileage may vary. Also, as of right now, I haven’t been awarded the CAREER grant (as I said, I just submitted my first attempt). While I did fashion my own submission based on these insights, I have no my idea if my specific proposal was any good. When I hear back, I’ll update this post to reflect the results and reviewer comments.

Anyway, this is what a successful SaTC proposal seems to look like:

Intro (~2 pages)

Describe objectives and demonstrate significance / impact. Frame and scope problem, discuss key insight, detail proposed activities to address problem.

• Employ statistics / figures to scope the problem and potential impact (e.g., number of people affected, economic impact). Forward-looking trends are common (e.g., year-over-year increases in something). Example scenarios are also common for systems-oriented proposals.
• I saw three framing strategies: (i) solving longstanding problems with existing tech; (ii) addressing burgeoning problems with emergent tech; and, (iii) solving a problem for an underserved population.
• Explicitly call out the intellectual traditions / disciplines that inform your work.
• Systematize the broader problem domain and discuss how your CAREER proposal fits in to this broader space (e.g., there are three big problems, this proposal solves problem 1)
• Bolded vision statement of proposed work
• Outline specific challenges that exist in current approaches / technologies that inhibit the vision (particularly if the work is scoped as being relevant to an emerging class of technology)
• Enumerate research questions and/or proposed activities
  — Can include associated impact statements with each RQ / activity
  — Can include explicit outputs & deliverables associated with each RQ / thrust
• Use a table / figure to provide an overview of proposed work and delta over prior work
• Kairos statement (why is this work important to do now?)

Background & related work (~1–2 pages)

Show that you’ve done your homework, that your proposal is different and new, and also that you’re standing on the shoulders of giants.

• Provide definitions of key terms from prior work.
• Use subsections related to each thread of prior work that you intend to build upon. Try not to be redundant with respect to the intro.
• Conclude each subsection / thread with a statement of contribution and/or integration.

Integrated ethos statements (~1–1.5 pages)

Show that you’ve taken some initial steps to reduce risk and demonstrate that you’re the right person to do the proposed work. This doesn’t have to (and probably shouldn’t) be in one single section, but can be interspersed throughout the document as appropriate. The high-level bit is to show that you’re equipped to handle this proposal as the PI.

• Detail any preliminary work you have done. This doesn’t necessarily need to be directly related to the proposed work as long as it informs the proposed work. The more related the better, particularly if there’s a risky assumption underlying your work.
• Showcase prior experience with proposed methodologies. Drawing from my own past experience, reviewers really care how well your skillsets — as demonstrated by prior / preliminary work—match up with the proposed activities.
• Demonstrate prior experience with related problem domains and technologies that you propose to employ.
• Highlight resources that might otherwise give you a competitive advantage (e.g., pre-existing expensive equipment)

Plan of work (~6–8 pages)

Deliver on the promises from your intro. Discuss, in detail, what you plan to do and how you plan to do it. This doesn’t have to be one section, though it often is. A common strategy is to structure your plan of work the way you laid out your high-level overview of proposed work in the intro (e.g., each RQ or proposed activity can be a subsection).

• Discuss how your proposed activities address your research questions. Sometimes RQs are first mentioned here instead of intro.
• Usually, the plan of work is divided into a number of different thrusts, each encapsulating a top-level contribution.
• Not all research thrusts need to be equally well described. Some can be more vague / loosely described than others. These are usually the ones that you will pursue later on in the proposal (because, presumably, things should change between years 1 and 5).

For proposed systems work:

• It seems like many different levels of granularity can be successful. You can include in-depth tech specs (including, e.g., system diagrams and lines of code) or high-level descriptions of the approach you intend to take.
• Example usage scenarios are helpful.
• “Old solution applied to new problem” style contribution: discuss the “old solution” in its original context and why it applies to this new problem space.
• “New solution to old problem” style contribution: discuss the “old problem” and why a new solution is needed / your key insight
• You can also propose to explore trade-offs between multiple novel systems / approaches if you are unsure which one might be best.
• You should have a technical evaluation plan with outlined and justified dimensions of interest.

For work that is contingent on the results of other proposed work in the proposal, you can provide examples of what you might do based on expected results.
For proposed user studies:

• Discuss how you plan on recruiting participants, particularly if you are targeting specialized user groups.
• Discuss analytic methodologies you might employ (e.g., iterative open coding).

For security work, you should discuss the threat model. If you have one threat model for all of the proposed work, then extract it out into top-level subsection. Otherwise, you can discuss the threat model for each proposed system / activity.

If you have multiple research thrusts, explain why you plan to pursue them in a specific order. Ideally, subsequent research thrusts will naturally build on earlier ones.

Educational activities and integration (~1 page)

Discuss your commitment to improving education for students both inside and outside of your home institution. Connect these activities to your plan of work— should be synergistic not independent.

Educational activities can either be extracted out into its own section or integrated with each research thrust. Common activities I saw include:

• Curriculum development. This can involve creating new courses or improving old ones. It can also involve creating MOOCs for public consumption.
• Writing a textbook and/or chapters of a textbook.
• Building educational tools (e.g., software to help students learn about specific topics)
• Workshops to disseminate research results to local community organizations and schools
• Academic workshops that bring together researchers
• Student mentorship and guidance

You may want these to be differentiated from “broader impacts” (which, for other proposals, may encompass some of these activities).

Intellectual merits: (~0.5–0.75 page)

Describe how your proposed work meaningfully extends the state-of-the-art in your discipline. This usually includes an enumerated list of intellectual contributions with an emphasis on novelty. Some possible contributions include:

• laying the foundation for a new field of study
• publicly released software / hardware artifacts and systems
• guidelines / design space recommendations
• improving emergent technologies to get them ready for primetime use

Broader impacts (~0.5 pages)

Describe how your proposed work impacts society beyond its academic contributions. Common strategies, here, include:

• Industry partnerships
• Impacts on specific communities of practice (e.g., developers, policymakers)
• Positive effects on marginalized populations
• Community outreach / K12 outreach
• Advising students, particularly those from under-represented backgrounds
• Disseminating results widely for public consumption (via non-academic channels like blog posts, social media, press)
• Running workshops on the topic at academic conferences

Again, you should differentiate what you write here from educational activities to avoid redundancy.

Timeline & project management (~0.5–0.75 page)

Demonstrate that you are a good science manager—that you’ve thought about the logistics of how to meet the vision you’ve outlined. It’s better to err on the side of promising too much than too little, if you have to choose (and you will you have to choose, because who can really predict what they will be doing five years from now?)

• Can talk a bit about personnel (e.g., graduate research assistants) and how their time will be utilized.
• Should cover both proposed research and educational activities
• The timeline usually seems to come in a list-based or tabular form. There are two types of tables: (i) rows for years and two columns corresponding to educational and research goals; (ii) rows for proposed activities and columns for years. The former is generally more space-efficient if your proposed activities cleanly fall along yearly boundaries; the latter is more expressive but less space efficient.

Prior support from NSF (~0.25–0.5 pages)

Show that you’ve been productive with any NSF awards you’ve worked on in the past. Kind of self-explanatory.

Letters of collaboration

Show that you’ve spoken to and motivated others who can help you execute the broader vision. I think all but one of the successful proposals I looked at had at least one letter of collaboration.

• Can be from organizational partners / other PIs with whom you propose to do activities. If you’re proposing to do field work somewhere difficult-to-access (e.g., a hospital), you should secure a letter of collaboration from one such place. Ditto for if you plan to work with a difficult-to-access population.
• It’s best when you can specifically outline what your collaborators will do: e.g., commit specific resources / equipment. Also best if your collaborators don’t demand exclusive ownership over IP.
• Take care to use standard template for these; otherwise, the whole proposal could be rejected without review.

[1] It is now the summer of my third year as an assistant professor. Many others attempt this sooner than I did — some people submit their first proposal in their first year. All this to say: definitely don’t use my timeline as a heuristic for when to attempt this 🙂

[2] Not that you need me to say this, but getting examples of prior successful proposals is, of course, step #1. I’ve found that most people are open to sharing if you ask nicely.

[3] Your Ph.D. advisor(s) and mentors in your institution may be good people to ask for feedback since they’re conflicted with you anyway and will not be allowed to review your proposal “for real.”

[4] Unfortunately, before the deadline I was pre-occupied with the deadline, so I was not able to put this post together. But, hopefully this will be helpful for people who submit in future years. I thought about deferring this post till next year, but the knowledge is fresh in my mind now so I thought it best to put it down on “paper.”

Thanks for reading. If you read this and thought: “whoah, definitely want to be spammed by that guy”, there are three ways to do it:

1. Follow me on Twitter (spam frequency: daily — weeklyish)
2. Subscribe to my YouTube channel (spam frequency: ¯\_(ツ)_/¯)
3. Subscribe to my mailing list (spam frequency: In Theory? Once a month with links to whatever I’ve written. In Practice? Once a year, when I feel bad.)

You also don’t have to do any of those things, and we will both be fine.