CS Grad School Admissions: Discussion with Prof. Ramesh Raskar (MIT)

I hosted this discussion organized for the volunteers of PathCheck Foundation who are applying for grad school in computer science in the US.

This blog is based on a discussion with Prof. Ramesh Raskar (MIT Media Lab) on successful graduate school applications to the US. The discussion focused on CS graduate school admissions, particularly for artificial intelligence, privacy, and computational health. You can find more such resources on Ramesh’s homepage or on his group website (Camera Culture).

Edit: I have added some points mentioned by Micheal J Black (Max-Plank) as well to the end of the blog!

The 7 Key Takeaways

1. The most important part of your application is creating a good portfolio of papers, blogs, videos, etc.
2. It helps to reach out to faculty beforehand especially if you have something meaningful to offer to their research — (i) Online platforms and conferences are great places to discuss their paper(s), extensions, and research. (ii) If you have something meaningful to offer, reach out to the faculty with a concise mail. It’s ideal to reach out to the grad students or post-docs and work your way up the system!
3. There are multiple ways to get noticed by a top school: publishing impactful papers in top venues; winning competitions and awards; extensive work with a top research faculty as an RA or intern (this could be a long-term process). This will not only build your profile for grad school but also help develop new relationships in academia.
4. While seeking grad students, faculty look for much more than just papers. They want to see soft skills, for example, time management, inclusivity, team dynamics, and the ability to build things (not necessarily production quality).
5. Since top schools get many strong and similar applications, any honors/awards/etc that can distinguish you will significantly strengthen your application.
6. It’s better to build a personal connection, and then apply to 10–15 schools, rather than applying all over the place.
7. Finally, there is always some level of randomness involved in the admissions process, and you should always keep that in mind!

Check out the below Q&A for a more detailed discussion. The answers are not verbatim, but I have tried to capture all the points Prof. Raskar had shared. We will be having another session open to the public soon. Stay tuned to know more details and join us on the call to discuss more on CS graduate applications to the US!

Q) How important are standardized tests, CGPA, and other “scores” in admissions?

Different universities have different thresholds for such things, but they are just a part of the application. Some schools are highly focused on such numbers but for most places, it would only be one of the factors in your application. You can think of test scores as a threshold: you will not be accepted to a school because you have stellar numbers, but poor numbers could be a cause of rejection.

Q) What counts as good research experience or research potential?

The best indicator of research experience is papers in top conferences, and to a lesser extent, papers in workshops. If you don’t have many high-quality papers at conferences, you can make it up by having a higher quantity of workshop papers. For example, instead of having 1 or 2 really good conference papers, you could work with 3 or 4 workshop papers. It’s not as great as having a top-tier conference publication, but if you have a few workshop papers, it could still put you in a good spot.

The admissions committee reads into more than just the content of the paper. Who are you publishing with? Are you able to create multiple papers with the same groups? This would show that you can work consistently with a team, you had strong teamwork skills that ensured prolonged success with a team. This would be evident if you have 2–3 papers in a row with the same set of people. Apart from that, the committee would also be interested to know if you write more broad papers or more narrow, impact-driven, etc.

There are other ways as well to show research potential. You could write some amazing blogs about existing works, commentaries, or review articles on the fields that you’re interested in. This can be a video or visualization or implementation etc of a recent paper from the group that you’re interested in. This might not be the strongest profile for, say, the top 5 schools, but this does put you in a good position to apply for the top 20 schools. It’s all about creating your own brand, as long as you have quality content on this! (Un) Fortunately, we live in a world where we are evaluated on a daily basis, for all good reasons.

Even if your paper is under review, it can still make a huge difference — especially if you have worked with really good people. It shows that you are working with someone and that led to this stage. You are coming to grad school to work with others, collaborate and create impactful research. If you could do all that on your own, you wouldn’t need to go to grad school. So what the committee looks for is more potential in these directions. The easiest method is to have a top-tier research paper, but all these — workshop papers, blogs, visualizations, implementation, etc. can still help your admissions.

Q) What do you look for in the letter of recommendation (LoRs), and do they need to be from your undergrad institution?

The letters are a product of 2 things — how good the letters are and the reputation of the person writing them. Most of the students would have the choice between a really famous person writing a brief letter and someone not so well known writing a deep letter. There’s no clear answer to which might be better, but in general, the reputation of the person writing the letter matters a lot. However, letters are mostly “necessary but not sufficient” parts of your application dossier. In short, it’s necessary that you have really good letters, but good letters alone may not help you to get into the top schools. Most of the time, the letters play an important role when your profile is already in the top x% after the first rounds of review based on your overall portfolio.

The recommendation letters will really help if they have something truly meaningful in them. For example, if the professor talks in-depth about the candidate and how she came from a foreign country and did an internship to build advanced software, a year later she came again (or the professor invited her back) — this is impactful. It helps to show that this is not a one-off incident but consistent and there is something special in the candidate that motivates the faculty to work with her more! While asking for recommendations, the candidates might have to share a draft or factual points with the faculty, nevertheless, the candidate must insist that the faculty add a few sentences that add color to the letter and can have an impact on the admissions. So always make sure you insist on this part with your letter writers!

Q) How important is work experience and how are candidates with work experience evaluated compared to freshers?

Work experience is a HUGE plus. This is especially useful, if you had some relevant work experience or if you were part of a complex setting (say a start-up or a new division that you were a founding member of, you worked in a really big org like Google and did something really important/interesting, etc). Extended work experience with one (or more) organization(s) will further boost your chances for a top graduate school. Say you work for 2–3 years or even 5 years and continue to do something amazing, that would be really helpful.

If you worked in a small team or in a start-up and you can describe these with research papers written, products built, and/or competitions participated in. Now if you have a letter from the CEO/CTO of this start-up talking about this, it shows you are someone who can make things happen. Selfishly, all professors are not just looking for smart people, but also smart people who can make things happen. Most professors’ research grants have deliverables — so they like people with strong professional skills (project management, discipline, time management, etc.). If all the candidate ever did was write something on the whiteboard and wrote 1 paper with them, it might not be what most professors are looking for. Working in the industry can help you acquire skills in these directions and help you manage your own time (which is one of the most important parts of grad school).

Q) Does having certifications, Coursera courses, executive education, etc. help to make your profile stand out?

If it’s something you can buy, then it’s probably not going to add much value. Also, it’s important how rigorous the course was. If this was say a semester-long intensive program with exams and selection criteria, then it could add some value. Let’s say you have a bunch of Coursera courses, it’s not a bad thing to have, especially if you want to show interest in a newer discipline. For instance, if you had an ECE background and later worked in the industry, maybe an AI in Healthcare course would help show that you took the time out to learn this new area and this could be a good indicator that you are curious and driven. For me, I always look at the candidates’ (privileged) background to gauge if they have this profile only because someone could pay for it. If that’s the case, then it won’t have huge value. This could be different if you want to get into say an MBA program.

Q) Should we reach out to faculty before the applications and let them know we are interested and applying?

It is a delicate matter. You can see on most of the forums faculty do mention don’t reach out to us, we won’t read your emails, etc. But in the end, we are all humans with our biases. As I mentioned before, all faculty are looking for really good candidates and to that extent, they would be happy to see a potential candidate reach out. That being said, don’t reach out to them if you don’t have anything meaningful to contribute (to their existing paper, or ongoing research directions). A more organic way of doing this would be reaching out to post-docs or graduate students in their group(s), initiating a discussion with them, and then later bubbling up from there. This could help you in 2 ways, 1) you get to know more about the research group than what you could ever understand from the websites, publications, etc. 2) it’s more likely that the faculty will see you as a valuable addition if you came by recommendation from a lab member. Again, the caveat still applies, don’t reach out to post-docs or graduate students if you don’t have anything meaningful to offer.

All this should be done with decorum: there is no point in spamming a faculty if you’re not getting a reply. Some faculty might find this a bit annoying. For me, I like candidates who take the initiative and try to reach out to me at events or on social media platforms to discuss my work or about something cool that they have done. So to answer the question, yes, it’s perfectly fine to reach out to faculty, but ensure that your email is not more than 3–5 sentences. You can get rid of all the fluff (esteemed university, group, it would be my honor, etc), Ideally, they should be able to learn about you from the notification bar from the first 2 sentences and the subject line). It’s important to work your way through the system. So try to contact the other group members first and then build a relationship. Sometimes these relationships take time to form and may take a good year or 2 as well! Additionally, closely go through the instructions given by each faculty on the email etiquette they expect before reaching out.

Q) How important is extra and/or co-curricular involvement for grad school admissions?

It depends on the activity. For undergrad admissions, we look for an all-around person but when it comes to grad school, it’s different. If this activity was maybe contributing to open source and building a community around it, then it might be useful, but if say you were in the basketball team — that’s good but, it’s not useful for grad school applications. For instance, I had a post-doc candidate who had brilliant ideas and we were interviewing him. After listening to him, I asked him to come back with more details on each of the ideas that he mentioned in a few days. Not only did he come back with a good write-up, but he also made some amazing sketches visualizing how AR/VR would be 10 years from now, this is how self-driving cars would be, etc. That made a huge impact! He later joined our group and helped organize a huge art expo for the entire MIT Media Lab community. In general, showing that you are an all-rounder might help, but it really depends on the group.

Q) How to write a really good SoP?

SoP is also a necessary but not sufficient part of your admissions. It’s important to have a good SoP, but an exceptional SoP alone may not get you into the top 5 schools. Many a time I use the SoP as a tie-breaker for say among the top 5 or top 10 candidates. This is what I do, there are some faculty members who give a lot of importance to the SoP and ensure to read every SoP in detail. The important part is, that you need to highlight the unique parts of you and what lead you to this research direction. For instance, you can maybe talk about your internship at a space agency and the algorithms they used for exoplanet viz changed the way you think about computer vision, etc. But in general, don’t use some template that you found online and make a few changes to it. The faculty can figure that out very quickly!

Q) How to show interest in a new field of research that you want to do in grad school?

You can’t just say that I worked on XYZ and now I want to do ABC which is not related at all. You will need to show some kind of track record in this new field of interest. This could be a Coursera class or a workshop paper or if you have written a review article about the field, a commentary on some of the recent papers, etc. Writing a commentary may be easy to get started, but it’s not enough. You can show that you have started something in it, but this may not be enough to get you into the top schools.

Also, don’t worry too much about such things. If you can’t get into a specific MS program you wanted, you can try to get into the same school in a different (but related!) MS program. Then you can maneuver your way through the school to join the original MS program. It’s always easier to get access to a prof/lab etc once you are at that school rather than when you are a complete outsider. Long story short, use your background in area “X” to get into a top program in area X (say top 10 schools) and then later use this as a launchpad to switch to area “Y” into a top 5 Ph.D. program! There are few programs in these top universities that may not be CS but are in the allied fields and are much easier to get into. You can use that opportunity to get into a top school and then network your way up.

Edit: Points from Micheal Black on the Linkedin Post from Ramesh
“In admissions, I’m trying to judge (1) ability to carry out research at a top-level, (2) passion for a problem, (3) a nice person who I want to spend 4–5 years working closely with and then be associated with for the rest of my career, (4) ability to overcome obstacles, (5) ability to write (so that I am not spending all my time as a copy editor). Critical things that help me understand all of that are (1) letters of reference, ideally from people who I know and respect. These letters should be specific and address the ability to do research. Letters that tell me that the person took a class and did well are worthless because I can get that from the transcripts. So the key is to do research as an undergrad and get letters from the research advisors. (2) Research statement. This is critical. It tells me how the applicant thinks, what drives them, and how they write. Poor writing here kills the application. I want to see some spark, some insight, here. (3) You don’t need papers at top conferences but it sure gets my attention. (4) What really gets my attention is when a colleague sends me an email saying “this student is applying and they are great” and explains why. This kind of champion is a gift.”

Follow this LinkedIn post to keep updated on any other advice: https://www.linkedin.com/posts/raskar_cs-grad-school-admissions-discussion-with-activity-6905556511515643904-qdsL

That concludes our discussion. Wish you all good luck with your admissions!

Thanks to everyone who attended the call and special mention to Govind Jeevan, Ramya B, and Sethuraman TV for copiously taking notes during the talk! Special thanks to Ramesh Raskar, Sethuraman TV and Sheshank Shankar for their help in proof-reading and editing the blog!

Please join our slack channel (PathCheck) if you want to work with a team of dynamic world-class researchers on computational health problems and build your research profile!