It would be remiss to discuss the NSF Fellowship without addressing Intellectual Merit and Broader Impacts. However, there is more than enough information on these criteria already available on the official website and elsewhere. Given this, I'll only discuss the strategies I used to address the criteria throughout my application.
Whether it accurately reflects your potential to advance knowledge or not, your academic record is important to your competitive edge. Grades, good and bad, are frequently mentioned on the Reviewer Feedback forms as a measure of Intellectual Merit. I've been told that the strongest applicants are often in the 3.5-4.0 GPA range. If you can, you should do whatever possible to raise your GPA as high as possible by the time you submit your application. It will be one less reason for a reviewer to count you out.
If you don't have a CV, go ahead and create one now (you're going to need it in the future anyway). Now, you don't actually submit a CV with the application, but FastLane does request information that would normally be listed in one. This includes teaching and work experience, scholarships and fellowships, academic honors, publications, presentations, etc.
List any teaching experiences, especially if STEM related; it doesn't have to be a formal Teaching Assistantship. List relevant internships and describe any research based work that you did. Academic honors such as Dean's List recognition are great. Mention all scholarships and fellowships earned, sometimes even honorable mentions (like NSF GRFP). Be sure to describe the selectivity (i.e. 10 recipients out of 2000 applicants or top 5% of department GPA) of lesser known awards.
If you have publications (in review or accepted for publication) in peer reviewed journals, provide the references. If you gave a presentation at a conference, in a poster symposium, as a live webinar, or even just for a class, list it, list it, list it! Technical presentations related to research that you've done establish Intellectual Merit. Professional development workshops that teach technical or soft skills that are essential to STEM leaders demonstrate Broader Impact.
You may not be able to write your own reference letters, but that doesn't mean you have no control over this part of your application. Find writers that know you fairly well as a person and a researcher. Letters from industry employers are fine, but they may not thoroughly address both criteria. Faculty members who have been awarded funding from NSF are good people to ask because they will already be familiar with the merit criteria. However, provide all of your reference writers with the tools they need to write a good recommendation anyway.
I'll keep this short because the next two posts are dedicated to the statements that you must write. Your statements are where you have the most flexible control over how you demonstrate the review criteria. No matter what, both statements should address both review criteria and the statements should complement one another.