Advice on writing an NSF CAREER proposal—or any proposal

• The people reading your grant will not be the same kind of people who would review one of your papers. They are from your field, but not from your subfield. That means, primarily, that they don't a priori buy into the significance of your problem. You must spend some time talking about how cool it would be to solve your problem and the cost of not solving your problem, in blood and money. (It also means that if you are proposing something in a trendy area with media attention, like 3D printing, you can probably get away with a bit less of this.)
• If you are proposing to solve problems in a new area for you—a problem other people have worked on but not yet yourself—assume that those other people will be reviewing your proposal. When writing a CAREER proposal, you are sometimes encouraged to propose work that can serve as a foundation for a research career and to distinguish your proposed work from your prior work. This can lead you to branch out into a somewhat new area for you. That is fine, but do your homework (read the literature, find a collaborator to write a letter). Someone in that area will be on the panel. If you aren't well versed in the literature, you may be dismissed for not understanding the challenges (and prior work) in that domain.
• Some additional successful strategies I have seen:
  • Intellectual merit: Transforming the field is a strong argument. Try to make it to others outside your field.
  • Proposal summaries are very important! They determine routing to program officer, panels, and reviewers.
  • Have evaluation plans and risk mitigation plan.
  • Pet peeve: People claiming to support a community (e.g. accessibility, age) with no connection to that community, or not evaluating with that community.
  • Have a compelling application focus.
  • CAREER wants project to serve as a foundation, not be something that could be done as an HCC small.
  • Ethics: intellectual property, inclusion, harm (can it make a gun)
  • Mention IRB somewhere if human subject research will take place.
  • Plan and budget for human subjects participation. Check the cover letter box.
  • Data Management plans get knocked for being short or generic. Make it long and thoughtful. Discuss kinds of data. Discuss user study data (mention IRB). Consider privacy of user study data.
  • Have a plan/letter of collaboration to evaluate your educational component.
• Here are section headings to consider:
  • Educational Objectives
  • Research Objectives
  • Motivation / Background / Research Gaps
• Each proposed research section ("aim" or "thrust") can have things like:
  • Background
  • Preliminary Work
  • Tasks
  • Recruiting Participants
  • Expected Outcomes
  • Risks (including a mitigation plan)
  • Evaluation Plan (inside each section or as its own section).
• Some of those are potentially quite short (a \paragraph{}, two bullet points).