Terms like Data Science (DS), Artificial Intelligence (AI), Deep Learning, and Machine Learning (ML) are more than just buzzwords in 2021. But, when it comes to MS CS/Data Science applications, there are many elements that might make the whole process overwhelming for the applicants — How does it feel to switch from MS track to PhD track? What do top universities look for in MS Computer Science and Data Science applications? How to improve your admission chances as a CS and Non-CS graduate?
Recently, I reached out to Dr. Arun Kumar (Associate Professor at the University of California, San Diego) to find out more insights from the perspective of a faculty member. He has been very kind to share advice and insights on what do top US universities look for in MS in CS/Data Science (and PhD) applications, how to improve admission chances and general advice for MS in USA aspirants. Additionally, he also walks us through his own graduate school experience as an international student in the US.
Advice on MS CS / Data Science Applications in USA
By Dr. Arun Kumar, Associate Professor, Computer Science and Engineering, Halicioglu Data Science Institute and HDSI Fellow, UC San Diego
My Graduate School Experience in the USA
MS in Computer Science Phase
I went to UW-Madison for an MS at first because I was not sure if I was cut out for research. I did some research projects with a few faculty and ended up publishing a top-tier paper in my first 2 years. All that gave me the confidence to switch to PhD and aim for a research career.
Challenges during the Initial Phase of PhD
Due to various circumstances, I had to switch thesis advisors–three times, no less!
So, my PhD trajectory was pretty non-linear, filled with all kinds of uncertainty, and not as productive as I had hoped it to be.
At one point in the middle, I seriously considered quitting my PhD. But thanks to support from family, friends, and all my advisors, as well as obtaining help from a cognitive-behavioral therapist, I decided to stay the course to finish my PhD.
Things Improved with Time
The second half of my PhD saw me becoming more productive in large part thanks to my terrific advisors, the awesome supportive environment of the Database Group at UW-Madison, as well as no-strings-attached funding for my PhD offered by Microsoft Jim Gray Systems Lab in Madison.
I started proposing original problems and ideas and executed them well, resulting in more top-tier publications that are now widely read. I also collaborated with folks in the software industry to help transition some of my research ideas to practice and established new professional networks via conferences.
Finally, I also came out of the closet during the second half of my PhD. That also helped boost my confidence and creativity. I have blogged publicly about my coming out experience.
Choosing a Research Career in the Academia
In the last 2 years of my PhD, I was weighing industrial research labs vs academia. I was fortunate to get to work with junior students (BS, MS) on extensions to my PhD research.
I enjoy the process of mentoring strong students and seeing them grow intellectually to produce new ideas. I also got the chance to teach the UG DB course at UW-Madison in my last year. I found teaching enjoyable too. Due to these reasons, I decided to go for an academic career.
Getting into Data Science
During my graduate studies, my research area was already a part of “Data Science” (DS) to begin with. My work is at the intersection of data management systems and machine learning. Both of these areas are key pillars of DS.
So, I was naturally inclined to be actively involved in the formation and promotion of HDSI at UC San Diego to help define and transform the future of Data Science research, education, and societal impact.
Advice on Careers in Computer and Data Sciences
How to Choose the Right MS Computer Science Specialization
Editor’s Note: There are so many excellent options to choose from including Software, Cybersecurity, Cloud Computing, Data Science / AI-ML, Information Systems.
All 5 of those areas in the above list are great specializations with good career scope in both industry and academia.
- My advice is to pick areas that excite you the most in terms of the intellectual content, the potential for impact on practice, and the day-to-day work involved.
- One way to find out if research work in an area excites you is to read some recent research papers from that area’s top conferences and think if you’d enjoy being in those authors’ shoes.
- If you want to pursue an industrial career, check out good courses in that area and see if you’d enjoy doing their programming assignments/homeworks.
Related Article: Top Computer Science Specializations for MS
MS Data Science: Hype vs Reality
If you just run after the next shiny thing, you will likely end up as more style than substance. That said, misreading genuine longer-term changes as fads could lead to costly lost opportunities.
Do your own research to assess if a change is longer-term or a fad. Two good signals I use to assess such things are the gradient and scope of said change across various stakeholders.
For instance, I bet back in 2016 that deep learning was going to be a massive change in ML and started a major research project that eventually became the bulk of my tenure case at UC San Diego.
Many senior faculty in computing, including world-famous ML experts, were still skeptical of deep learning in 2016. But I saw that many domain scientists, enterprises, Web, other companies were excited about deep learning’s potential to unlock unstructured data for analytics. I am glad my bet panned out well.
Likewise for DS as a new discipline, UC San Diego too bet big by launching HDSI in 2018. Many universities were skeptical (some still are).
But in the last 3 years, we have attracted top-notch faculty to HDSI, with some turning down many top 10 departments in CS and statistics! In the last two years, more top schools in the US have launched DS programs.
The momentum is only growing, not just in academia, but also in industry in terms of the kinds of jobs that DS expertise can lead to.
Related Article: When an MS Data Science Degree is Not Worth it and Top Alternatives
Choosing the Right University for MS Data Science
I’d be wary of universities that simply tack on some CS courses to a statistics degree or vice versa to jump on the bandwagon without deeper pedagogical thought on the curriculum.
Look for programs that are carefully designed to achieve a true synthesis of the underlying disciplines to train “Data Science-native” thinkers.
This is what we have done/are doing at UC San Diego HDSI with our BS in Data Science. For a hands-on experience, all of our DS majors work on a capstone project in the final year.
HDSI also just launched our PhD, MS, and Online Masters degree programs in Data Science, with applications due Jan 15, 2022. I see this as a paradigm shift in the future of Data Science in US academia in general.
Graduate School (MS & PhD) Application Tips
How to Shortlist Universities for MS and PhD in CS/DS?
Choosing the Right School for MS
For MS, my top criteria would be;
- the brand name of the school
- breadth of the program’s course offerings
- potential for or actually promised funding in that department
- quality of life in that city–in that order
Note that we are now in an “era of plenty” for CS/DS jobs. So, an MS from any of the top 40-50 schools is pretty reasonable for industrial careers, although an MS from a more well regarded school is likely to lead to a slightly higher starting salary in your first job after MS.
Choosing the Right School for PhD
For PhD (or MS as a stepping stone to PhD), my top criteria would be:
- a strong research advisor who is interested in working with you and has funding
- presence of more faculty in that area or nearby areas (not just the advisor)
- quality of life in that city
- and school brand name–in that order
Don’t Focus Too Much on Brand Name or Rankings for PhD
It is often a surprise to students from India that the criteria for PhD are so different from those for MS. In particular, note that school brand name or US News rankings are all effectively irrelevant for predicting PhD success after you condition on the strength of the advisor and the advisor-advisee fit.
I have seen far too many cases of international students being naively enamored by brand name or rankings, ending up with a poor advisor fit, and ultimately resulting in mediocre PhD outcomes.
Related Article: How to Shortlist Universities for MS in USA for Fall 2022?
What do Faculty Members look for in Applicants: My Perspectives
I do not recruit external MS applicants; most faculty do not either. I sometimes recruit MS students who have already joined UCSD into my research group.
For external PhD applicants, I look for a strong fit on research interests, evidence of research potential (ideally, relevant prior published research), strong recommendation letters, and a solid SoP. I then create a shortlist for virtual interviews and then decide who to make offers to.
How do US Universities Evaluate MS Data Science Applications?
For an MS in DS, most US universities admit based on the decision of a committee with multiple faculty. Primarily they look for relevant technical background, good grades in UG courses, good SoP, strong recommendation letters, and sometimes, evidence of initiative (e.g., projects, extracurriculars, etc.).
Good GRE and TOEFL scores help a bit but bad scores can hurt the case. The brand name of the UG school can help but it is usually not as important as the other factors above.
My Advice for MS CS/DS Applicants with Average/Low GPA and/or GRE
Do not obsess over only a small set of schools. Apply widely to spread your eggs across many baskets, subject to the brand name and quality of the school’s education/research you are comfortable with. Be realistic in your expectations.
The level of competition in both CS and DS is intense. For instance, CS is now the most popular major at most US universities. DS is usually second in the schools that have it.
Do Online Courses / Certifications Help in Applications?
The top 25 US universities do not give much emphasis on this because they have so many CS major applicants already. I am though not sure about lower-ranked universities.
How should non-CS graduates prepare themselves for MS CS / DS applications?
- UG in a CS-adjacent discipline such as ECE/EE, Math, Statistics, or Data Science should generally find it not much harder to get into CS MS programs than CS majors.
- But if your UG was in a totally different field (e.g., chemistry), the chances are lower. In such cases, it can help your application, at least in some universities, to show those online courses/certificates in CS.
- It will also help to have relevant CS-related project experiences and/or those programming contests to show your initiative.
Most Common Mistakes Among Indian and International Applicants
Most faculty are heavily time-crunched. SOPs that start with flowery statements, tangents about one’s childhood, etc. waste most faculty’s time. Make it easier to evaluate your case in a quick skim read. Get quickly to the point on why you want to do that MS program and where you want to go in your career.
If some of your UG grades are poor and/or there are other anomalies in your record, make sure to explain them clearly in your SOP, e.g., unusual economic or social circumstances that caused your grades to dip. Brushing such things under the carpet is almost always seen as a red flag.
Related Article: What do Admission Committees Look For in SoP / Personal Statement?
How should candidates with career gaps (specifically, female candidates with maternity breaks) approach their graduate school applications?
Just mention it in your SoP. Almost all faculty understand that maternity leaves and even paternity leaves are a normal part of life. These are not considered disadvantages for your case. If there is some other reason for the gap, explain it honestly in your SoP.
Upcoming trends in Data Science and AI/ML that students should be aware of
In core ML/AI algorithmics, there is much focus on reducing the bloat of deep learning models with more domain-specific invariants such as physical laws in spatiotemporal analytics and text semantics in NLP.
In ML/AI theory and statistics, there is much focus on rigorously explaining why deep learning methods work so well in practice, quantifying their robustness and security, and ensuring Data Science methods do not cause discriminatory impact across various groups of people.
Finally, building systems and data management tools for ML/AI workloads, data-centric tasks in ML/AI, end-to-end ML platforms, MLOps, and cloud-native tooling for ML/AI are all hot areas in both research and the software industry.
General Advice for International Students on Pursuing MS in USA
- Find friends to share your apartment/house with because rents are high in most US cities. If they are in the same MS program, you can study together, exchange notes, and grumble about the same faculty!
- Sign up for relevant TAships. Do not be disheartened if you are not able to secure an RAship–the vast majority of MS students will not either, both due to the level of competition in CS/DS and because most faculty prefer PhD students for RAships. Look out for project assistantships too or RAships in other departments, e.g., applying your CS/DS skills to domain sciences.
- Apply for industrial internships for your first summer to get practical experience in the US software industry. Networking and extra income, especially if your MS is funded by a loan, are other key benefits of industrial internships. Apply and prepare for interviews by the end of the Fall semester/quarter itself.
- Learn to cook, do your own laundry, and clean your apartment. Eating out all the time will eat into your savings. Do not ignore your health, both physical health and mental health.
- Do not work/study all the time. It could lead to burn out. Take sufficient breaks on weekends, vacation periods, etc. Travel to other cities and states if you can. The US is a huge, varied, and beautiful country!