from Guide to Undergraduate on Oct 9, 2022

How to get into research, as an undergraduate.

I've received many many dozens of emails from undergraduates looking for research, during my time as a PhD student; here's the post I wish all of them read.

Looking for research? I've been on both sides of the table — both looking for and providing research opportunities. It's frustrating as an undergraduate, but funnily enough, it's tough to recruit as a PhD student too. So, I'm writing to you research-bound undergraduates: There are researchers out there that need your help and are having a hard time connecting with you. To stand out, focus on communicating your talents effectively, because this short ~150-word pitch — the cold email itself — is the biggest weakness I see.

If you're reading this post a bit farther in advance, you may be looking for longer-term advice for readiness. In which case, check out courses and other ways to prepare yourself for research in How to prepare for research, without prior experience.

99% of opportunities aren't formal applications.

This post is tailored for "ad-hoc" research collaborations, where the path to collaboration is nebulous. This forms the bulk of research opportunities. However, for sake of completeness, I'll mention the few examples of more structured research programs, where it's obvious how to apply and get involved:

There are several summer research internship programs organized nation-wide by the National Science Foundation, called Research Experiences for Undergraduates (REUs, link) at different institutions. Universities will have specific programs too; UC Berkeley for example runs 3 Summer Undergraduate Research Fellowships (SURF, link).
Organized year-round research opportunities also exist. You can find them at 3 levels of specificity: organized by the campus, a lab, or a single faculty member. At UC Berkeley, the campus runs an Undergraduate Research Apprenticeship Program (URAP). Several labs (lab directory) have specific application portals — SkyLab uses DARE, for example. Even more specific still, some professors have their own processes, and in most cases, they list requirements on their websites.

That's it for the more structured programs; for what it's worth, some these applications are low ROI — many applicants, few positions, and possibly out-of-date project descriptions. I'm going to take the rest of this post to instead focus on the 99% of opportunities that exist outside of formal applications.

Send PhD students short emails.

There are 3 very quick, short tips I can offer that most undergraduates don't follow:

Reach out to PhD students, not professors. The former know which projects need help and in what way. They also have more time than professors to respond to emails.
Express interest in a field that the PhD student publishes in. Skim the PhD student's website and see which papers they've just published or are most proud of.
Keep your email short. No essays!

Provided you're short on time and sending cold emails now, this email is the only part of the process you control, so make it count. There are three possible states your reader is in:

Your reader thinks they don't want to work with an undergraduate collaborator. Most likely, they've never worked with an undergraduate collaborator before.
Your reader wants to work with an undergraduate collaborator but doesn't know how to. Afraid to take on mentees, when they themselves don't know what's going on.
Your reader wants to work with an undergraduate collaborator but doesn't know how to "filter" good candidates. Heck, they don't even know what "good" candidate means.

In short, you need to convince a reader who a) doesn't know they need undergraduate collaborators, b) doesn't know what they're looking for, and c) doesn't know how to collaborate with you. Your email has to address these 3 items in some way, and here are the basic 3 points you need to demonstrate. Whether or not the PhD student knows it, this is what they need:

Readiness — Show you're technically prepared for research, with the right coursework and maybe even experience beyond coursework. Some academics care about grades, but I daresay more important than grades, is your ability to meaningfully hold a conversation about the topic.
Motivation — Show you're broadly interested in the same problems. If you have an extra reason to be motivated, include it. For example, if you want to apply to PhD programs. No need for larger-than-life missions or a personal history that has predestined you for this moment. Just interest in the field is good.
Work Ethic — Show you're hard-working and have the time to commit. For the latter, "reduced" unit count and lack of other time-consuming extracurriculars works. I put "reduced" in quotes because it's all relative. If you normally take 20 units and you're taking 10 now, great. If you normally take 10 units and are now taking 5, even better.

Include readiness, motivation, and work ethic in your email.

Your goal is to communicate how invaluable of a research asset you'll be, in as few words as possible. Please, no essays! It's fairly difficult to communicate your value over an email but the below is the bare minimum. For each section, you can find example blurbs. I also included a version that reflects my junior-year experience:

Interest (2-3 sentences)

Show that you know at least which field this PhD student works in. Anything to suggest you at least looked at their recent publication titles.
The second-most important piece of information here is your year, so in other words, how many years you could potentially conduct research, before graduating. It takes time to ramp up, so a last-semester senior will likely not be productive for either party to collaborate.

Readiness (2-3 sentences)

Most important: Mention previous research projects. Include publication venue, authorship, your role, and names of PhD student and professor. If there are more than one project, mention just the faculty names for the others. Leave descriptions for your resume. The goal is to show previous successes or at least exposure.
List abbreviated course titles. PhD students are likely from a different university so don't know what course numbers mean. Include your grades only if they're good. For UC Berkeley students, remember our GPAs are deflated, so an A- is not as well appreciated. I would definitely not include my own grades.
Depending on the PhD student, industry experience could be a boon or a burden. List companies and achievements if nothing else. Effectively, show you can code and know the process of coding with others (e.g., how to submit pull requests).
If you still have room, cover any relevant prerequisites that you've satisfied. For BAIR in particular, mention knowing PyTorch / Tensorflow and in what context you worked with those frameworks. For Skylab, knowing how to build serverless applications might be worth mentioning.

Motivation (2-3 sentences)

If you're interested in applying for PhD programs, definitely mention it. If you're interested in applying for the 5th years masters program, definitely mention it. If neither, don't mention it. We're primarily interested in PhD applicants as we know there's a baseline level of commitment.
List how many hours you're willing to commit and explicitly mentioning in-person in the lab is a bonus. Mention other commitments you're cutting to make this happen, to make it more convincing. We know the verbal hour commitment is always overstated.

Make it easy. In short, should require as little work as possible for them to say yes. That's it! Here's the full draft, for easy copy-and-paste.

Hello researcherJohn Doe,

I'm a year3rd-year undergraduate majoring in majorComputer Science and am interested in fieldcomputer vision, specifically in your subfieldneural architecture search work. Do you have research opportunities for undergraduates?

I've previously worked with collaboratorVlad Feinberg from labProf. Gonzalez's lab on topicexpanded-length rollouts for RL, which we submitted to conferenceNeurIPS 2016 (second author). I contributionsran ablations and helped frame the story. For preparation, I took courses in coursesmachine learning, optimization, and probability theory; and prereqsinterned at a startup for coding experience.

I hope to apply to programPhD programs in timelineFall 2017. Knowing that, I've put aside other commitments to dedicate myself to research. I've [commitment 1: reduced courseload from 20 to 10 units], [commitment 2: dropped all extracurriculars], and can dedicate hours20 hours per week, working in-person for days3 days each week.

I've attached my resume and am happy to provide additional information. If helpful, I'm available availabilityanytime this or next week after 3p and if you let me know a preferred time, I can setup the gcal and Zoom.

Thanks, [name: Alvin]

Prepare for research.

Here are some more tips depending on your responses:

Got a positive response from your email? Congratulations! Your next step is to see How to succeed as a (research) mentee. . This is of course how you'll maximize your collaboration.
If you've been rejected or didn't get responses, not a problem. You may want to see How to prepare for research, without prior experience.

Finally, regardless of the outcome, your next meta-goal should be to understand what you can get out of research. In particular, see What defines a "good" researcher?

← back to Guide to Undergraduate