An in-depth post on my experience deciding to do my PhD and the application process. I recently accepted my offer, so I decided to write this while everything was still fresh on my mind. My thoughts may change over time. I write both about my personal experience and advice for future applicants. I ended up getting 6 acceptances from (the obsurdly high) 19 applications I submitted. I knew I wanted to go to grad school this cycle and had no idea how my application would be recieved so I applied to a lot! I do want to disclose that I am a domestic student, a woman, and went to a prestigious university for undergrad. These are all factors you probably cannot change at the time of applying to PhD programs but may alter application outcomes.

Deciding to Apply (March 2023)

I was working full-time, almost 2 years out of college when I decided something needed to change. I loved my job and the research I was doing at the bank, but the amount of red tape for doing my job was increasing drastically (thanks ChatGPT). I wanted to be in an environment that allowed me more freedom and flexibility, as well as focus only on research. When I started really thinking about the career I wanted, it became more and more clear that I needed a PhD. However, I had never worked in an academic research lab and didn’t think I could actually get into a program. I sat down with my mentor from undergrad and told him I thought I should go get my masters first. He pushed me to do the PhD applications instead, reminding me that 3 years of full-time research and my patent outputs were quite a valuable experience. I want to thank David Mimno for his advice and confidence - I wouldn’t have gotten into a PhD without his guidance.

I think this is a very personal decision and you truly have to be ready to spend 5 years doing research. My advice would be to think about what you want to do in the long-run and work backwords. What steps do you need to take now to get to that point? Obviously, people like the research they are doing but its still work. I don’t believe the people who say you should only do a PhD if you are so passionate you couldn’t think about ever not doing it! I think its better to ask, can I commit to the oppurtunity cost of graduate school? I have worked for 3 years and I think there are a lot of upsides of working first. I feel more commited in my decision, I know I’m not just doing it because it’s the next logical step in my education. I am in a much better financial position than most my peers will be, since I have money sitting in retirement and investment funds that can grow over the next 5 years! For reference, money invested at a young age is worth the most. If you maxed out your Roth IRA at 22, that $7000 would turn into about $96,000 by the time you retire. If you start at 28, it will only be worth $67,000. At 30, it’s only worth $60,000. So, building a bit of a nest egg, paying off any student debt, and getting a retirement account started may be very worth it in the long run and not something people often talk about.

Pre-Application Research (April - July 2023)

I decided I wanted to get my PhD in March of 2023, right around the time the previous cycle’s admits were coming out. While I was sad I had missed that cycle, this turned out to be one of the best advantages I had. I knew I had about a year to research programs, do my applications, prep for interviews and learn how to choose which program was right for me. This mostly consited of identifying programs and faculty, reading as many papers as I could, and cold emailing people.

Identifying Programs:

The first step I took to decide where to apply was to think geographically. Currently, I am in Dallas, Texas but my close family is in the upstate NY area. Being closer to them was really important to me. I created a rule that programs I applied to had to be either (ideally both) less than a 6-hour drive or a direct flight to Rochester (exceptions were made for Stanford and UW). This left mostly schools in the Northeast and mid-atlantic. I also know that the price of applying is important to people. Kaiser Sun has created a great resource for which schools have fee waivers & who is eligible. I was able to get fee waivers for more than half the programs I applied to, which allowed me to apply more broadly without going over budget. I also did use csrankings and open rankings to generally get a feel for how much NLP research was happening and the general reputations of universities. If a school met these general criteria, I moved on to looking at the schools specific labs.

Identifying Faculty: For me to apply to a school, I wanted there to be at least 2 professors who I would be happy to work with. This is because there are a lot of things that happen in life that we can’t expect. Maybe your first choice isn’t hiring that year, or you commit to a lab and then it doesn’t work out with the original PI. You want to have a backup! There are a couple things that wereare most helpful. The first is getting an understanding of which professors work at what universities and what they work on. CSRankings is particularly good for this, as it links professors websites, google scholar progiles, and more. Be careful to note CS Rankings sometimes misidentifies people so the lists may not be the most updated or correct. Find out who the key figures in the subtopic you want to work on are. I spent a lot of time looking at papers I liked, and then going to look at those authors labs. You can look at who they typically collaborate with as well! Also, remember to look at program’s pages for different research groups and the most updated faculty lists.

Reading Papers: This is the most important time to read as much research as possible. Before you start your SOP or applications, before you have to do interviews, before you need to even know exactly what you want to do. Get as well versed with the main papers in your subfield as well as new research to be up-to-date. When I say you will thank me, you will thank me! I know that you may already by reading a lot of papers for your current job or degree, but ramp it up. Double whatever you are currently doing. This will help you not only identify who the important people and instituations are, but also will make your applications and interviews go much more smoothly. I recommend doing this from May until September (then you need to pivot to applications).

Building a Support Network: You need to build a group of people who can help guide you through the application process. This means people who have done it before! I found it helpful to start getting in touch with my mentor from undergrad and a friend I had from undergrad who is currently a 3rd year NLP PhD student. They were able to give me some early advice, and put me in contact with other people who would be good to talk to. It’s helpful to get advice from both students who have recently gone through applications as well as professors who are familiar with finding potential candidates.

Cold Emailing: Tread carefully here. All in all, I think it is helpful to email professors, but it has to be done correctly. If you are not very confident in your interpersonal skills or English proficiency, you may be better of not emailing. Of my 6 acceptances, I had emailed 2 of the professors. If you decide you would like to, here is what I did:

  1. Check professors site to ensure they do not have a ‘no emails’ policy. Do not email someone who has this. It will just piss them off. Some professors also have specific instructions, like a key word or form they would like you to fill out. Follow all instructions!

  2. Read one of their recent papers and generate a few questions. You want to find a paper from the last year or two because you want to talk about a project the Professor is still interested in! You want to do this to include later on in your email - it isn’t necessary, but I think doing so will get you much higher response rates.

  3. Draft the email. I made a template that I changed for each professor. Google docs has an email template, so I drafted emails there and saved them in a folder. This was super helpful when I was doing applications later on because I had already verbalized why I wanted to work with a professor and could just reference that later on. My general format was an intro about myself, a paragraph about our research match, and a conclusion with a personal connection. You want to show them, briefly, why they should want to work with you. That means 2 things: you do cool research and you are a cool person. People often forget the human aspect! One of my potential advisors said my email was the only one he got that he knew wasn’t written by chatgpt because I mentioned his french toast. End the email with something you have in common - maybe a shared alma matter, hobby, or just pay them a compliment.

  4. Send the email but make a catchy subject line. I went with “Cornell Grad Excited About Your NLP Research”. You want something that compliments them and tells them that you are legit. Send from an academic email so its more likely to get through any email filters a professor has on.

Example Email:

Dear Professor _______,

My name is Jenna Russell, I’m a 2021 Cornell graduate and I’ve been a research data scientist at Bank of America (BofA) in Plano, TX since leaving school, working on our virtual assistant Erica. I’ve realized how much I love the research aspect of my job and am applying to PhD programs for Fall 2024.

At BofA, my research has centered around projects such as improving Erica intent predictions and translating Spanish to English. We also do a lot of work improving the experience for customer service agents. This year I’ve primarily researched creating dialogue summarization systems and creating a retrieval-augmented QA system for agents to use. I really enjoyed reading [paper name] earlier this year. I liked how [something you liked]. [Question about paper] ?

I believe you would be a great advisor due to our mutual interest in [mutual interest]. I also really liked your food pictures on your website, it looks like you make a mean french toast.

Best,
Jenna Russell

Letters of Rec (August - October 2023)

I started working on my letters of recommendation in late summer. I had already been in contact with my letter writers. My three recommenders were my advisor from Cornell, my current team lead at BofA, and coworker (who used to be a stats prof at UIUC). I think that 2 LOR need to be academia insiders, and one needs to be a known and hopefully well-liked professor in your field. I will say I think having connections from Cornell really did help my application get seen.

The best advice I found was to make it as easy as possible for your recommender to write the best letter possible. I made a pdf detailing all the schools and fellowships I was applying to, with deadlines. I included all information about past teaching and researching projects. I wrote bullet point lists about all projects, my contributions, and the results of those projects. This made it really easy to keep track of everything. I also sent polite emails whenever there was a deadline nearing (at least a week notice). I also kept a spreadsheet where I monitored which of my LOR writers had submitted to which schools/portals. This was really helpful for my sanity, as well as being able to remind them which schools were due when. Also remember to thank them and give them a handwritten note of appreciation after application season is over!

Application Materials (August - December 2023)

Being Organized: August was when I really started honing in on my application materials. This meant getting a final list of schools, starting my application materials, and getting organized. Being organized is key! I had spreadsheets with a lot of information! Keep track of all the schools you are applying due, key due dates, passwords to all the application portals, and more. One of the biggest tips I have is to start filling out the applications early. You can get the boring stuff overwith early on and make sure you aren’t surprised by any requirements. I kept track of which schools needed which essays and what special requirements they needed. This was helpful, because usually if one school needed something another school did too. In the end, I had saved 1000 and 2 page versions of my SOP. 250, 500, and 1000 word versions of a diversity statement. I was able to rework a lot of the same material to fit requirements for multiple things.

Statement of Purpose Start writing the statement of purpose early! I started my first draft in August. I know this was super early, but I was glad I had so much time to rework it create many rounds of drafts. I wrote the first version and got some good advice by early September, but I knew it wasn’t exactly what I wanted. I wanted to truly convey why I wanted to do specific research and create a narrative connecting my current work to that work. I stepped back and spent a few weeks reading as many SOPs as possible - the site notion is very helpful for examples of past successful essays. I also tried to find and read blogs other PhD students had written. I was able to get to work rewriting and felt much more confident in my message the second time through. I did utilize AI for some general editing ideas, but I made sure I had a clear idea of the point I wanted to get across. Mostly I used for paraphrasing some small phrases to be more succinct.

The best advice I have is to get as many people to read your SOP as possible. Especially people who know the subfield and what a good SOP looks like. I asked my teammates (all of whom had PhDs), the friends I knew who were already in programs, and mentors from applicant support programs. A ton of schools had programs to help underrepresented groups in CS get some feedback on their applications. I got mentors from Cornell, UW, CMU, and MIT, most of whom were current students. One of my mentors was in my subfield and was also able to give me some advice on which professors would be good to work with and sent me materials like good blogs and info to use. I was able to have many rounds of revisions, both because I had so many sets of eyes looking at the work and because I gave myself so much time. I never felt like I was scrambling.

I think you should take advantage of the header of the SOP. This is a chance to include some information about yourself to make your profile shine. Under my title, I had a line that included my academic email and a link to my website. This was an easy way to help people reading your SOP make a mental picture of who you are and make you more memorable.

Supplimental Materials: I did also write/revise my supplimental statements, but honestly I put a lot less effort into these. I felt it was pretty easy to write since I had such a history of advocating for women and URM in CS/Data Science spaces. I also think most advisors don’t put as much emphasis on this. As for my CV, I had a few current students make some edits on this, which was also helpful! Big advice was to keep it to 2 pages.

Website: Make a website! It is 2024! Github pages has a very friendly interface that someone who is signing up to do a CS PhD should be able to figure out pretty easily. You need to have a nice homepage with an intro to who you are. Put any publications/relevant projects/cv on the page. Lastly, put an about me page!!! Personally, I think the human aspect of applying to PhD programs is highly undervalued. In the end, your materials will get you to an interview, but your personality has to shine as well. Show that you are a person with real interests outside of work. A PI has to sign up to want to work with you for 5+ years.

Interview Process (January - March 2024)

One thing that was helpful was preparing for the interview. For each one, I made sure to research the professor and the projects they were working on recently. I made sure to read & annotate one or two papers where my interests intersected with the work and create some questions/points I could bring up during the interview. This was helpful in terms of painting a picture for the interviewer that I would fit into their lab easily. I also made sure to know a few recent works in the field that were interesting so that if asked about a favorite paper (this happened a lot) I would have a relevant paper to the field that would also show I was on top of my reading!

Interviews varied greatly by professor. The vast majority were short ~30 minute conversations where I was asked things like why I wanted a PhD, my greatest research interests, and what my research experience had been thus far. I did have a few that asked some more technical questions but even those weren’t too bad. Honestly, if you get to the interview stage of the interview - you’ve already qualified for the spot. A professor isn’t going to waste time interviewing people that aren’t a good bet. Most likely, this is where a professor is figuring out if you are someone they can work with for the next 5-6 years. While not the most helpful advice, the best thing you can do is pretend like the professor is your friend. Speak concisely, be yourself, and breathe. This is also a good time to get a feel for if you would like to work with a professor. If it seems like there isn’t much to talk about, just ask questions. Ask how the professor likes the program or what they’re excited about working on or even if they like the city the school is located in. Overall, the interviews were less stressful than I anticipated.

Visit Days (March-April 2024)

Visit days were super helpful in terms of understanding lab dynamics and see what the ‘reality’ of joining a program will be. It also is a great networking oppurtunity - academia is small! These are peers you will likely see at conferences, internships, or may be good to know later on. Attend all visit days you feasibly can (obviously if you know you are 0% choosing a school, let the professor know so they can have someone seriously considering that program go visit). Make sure to get time alone with the students in the labs your are visiting. They are usually pretty honest about their relationships with their advisors and their work life balance. This definitely helped me rule out one program.

Choosing a PhD Program (April 2024)

This was a really hard decision. I had an option (Penn) in a location I knew I would love (Philly) with a lab that seemed super supportive and social, but the professor was more in the ML world than the NLP world. I had an offer at an amazing NLP group (JHU CLSP) with a well-established prof and social lab, but I knew I wanted an advisor that had more hands on time. In the end, I chose UMass even though it was the least ideal location of the three, mostly because of my close research and personality fit with my advisor. I also knew I could always do internships in different cities or spend time elsewhere later on in life. It didn’t hurt that UMass was a quick train ride from NYC, drive from Boston, and a 4.5 hour drive home for me. I also grew up in the snow, so I am happy about the easy access to skiing. I think if I didn’t have a car, have family in the northeast, or actively like snow, this would be a harder decision for me. I was super undecided until the last minute, but as soon as I choose I was very happy with my choice :) All you can do is go with your gut.