Managing Undergraduate Research

10 minute read


Hello! This post contains suggestions to undergraduates who are interested in research or are willing to find out if some form of research interests them or not. While the article is based on my experience at IIT Bombay, I believe that most of my suggestions will be helpful to most undergraduates.


Research is a process of investigating systematically. Universities across the world are judged based on their quality of research. One of the first things that you, the reader, might be interested to find out is, if academic research interests you or not. One of the best ways to find out is by actually doing research and preferably with a professor.

What is the right time to talk to a professor? How should I approach a professor? I don’t know a lot. I had similar questions and fears till I actually plunged into solving research problems in my third year. In hindsight, I shouldn’t have hesitated so much.

Contacting professors

Most of the professors would be willing to guide you if you show interest. Before you contact a professor, read at least one paper or listen to a talk they gave in the department. It is okay if you don’t understand anything beyond the introduction. Spend some time reading a tutorial on the area of the problem. Now that you’ve done your homework, you’re in a slightly better position to judge if you really liked the professor’s work. Now write a short email indicating your interest and the homework that you’ve already done. Write a subject in the following manner: “Request for a research project, student from IITB.”

You’re much more likely to get a response to these emails compared to half-hearted template emails. However note that you might still not get a reply. Remember that professors are very busy. After a few days of not getting a response, reply on the same thread politely “Professor, I understand that you must be busy but I’ll be obliged if you could let me know if there are openings in your group.”

Being polite and following an email etiquette should become a habit. Read more about it at this blog post by Prof. Matt Might.

Discomfort of research

Every researcher has to get comfortable with a few realities of research:

  1. There will be no one to spoon feed you and there are no regular exams to keep a check. This may feel liberating but it comes at a cost of being responsible.
  2. Being a weekend warrior simply doesn’t get things done. Irrespective of how efficient you are in clearing courses with one-nighters, research doesn’t progress without consistency.
  3. Progress is typically slow and happens in spurts. One has to get comfortable with being uncomfortable.
  4. Your advisor will not know every detail of your work. If they did, they would’ve already published the work. Research is much more collaborative and requires active participation.

Working with an advisor

Advisors like the students are human beings who come with their own strengths and flaws. It is up to the students to develop a strong professional relationship with the advisor and utilize the advisor’s experience and expertise to propel a project ahead.

You must meet your advisor at least once a week irrespective of the “progress” you’ve made. It is better to report that nothing worked rather than reporting nothing at all. Always remember that failure is information and information pushes your project ahead. Have a discussion with your advisor about what and why things didn’t work. The discussions generally turn out to be fruitful and give enthusiasm to try new approaches.

Many advisors ask undergraduates to work with PhD or Master students. I am personally not very fond of such arrangements. However, such arrangements are unavoidable in many circumstances. Try to learn as much as you can and take responsibility for the work. Always remember that the success of the project is going to help you and not doing enough will only lead to regret later.


There are a couple of skills that every undergraduate should pick up:

  1. Spend time learning LaTeX. If you’ll ever need to typeset math, you won’t survive without it.
  2. Learn to deliver presentations. Here is an excellent guide for giving talks by Prof. Jonathan Shewchuk that I have also found incredibly helpful.
  3. In my opinion it is better to use Powerpoint instead of Beamer. Use IguanaTex to insert TeX in Powerpoint.
  4. Learn to read research papers
  5. Develop some technical writing practice. I suck at it. Following Prof. Matt Might’s advice, I will try to regularly blog about ideas I find interesting.

Which area to work in?

Part 1: Request to diversify AI research

Do what you’re passionate about! That’s the most hollow advice I can give you because it is not realistic. Market forces play a big role in what we end up doing and there is nothing wrong in accepting that. However, if everyone ends up doing the same thing, then it creates an unhealthy environment for everybody. You would accuse me of blasphemy and remind me that AI is the new electricity. I couldn’t agree with you more but most of the students end up working on limited areas of AI research, i.e., pattern recognition problems arising in Computer Vision, Natural Language Processing, Speech Recognition and Medical Computing. I mean no disrespect to people working in these areas and I’ve immense respect for my colleagues and my seniors who have done top quality work in these fields. My problem is that so many talented folks are working on very similar problems. Read what Claude Shannon had to say about Information Theory when the theory was considered as the new electricity.

I feel that there is a serious need to diversify. I am aware of the high that one gets on training a neural net classifier to distinguish between cats and dogs. I don’t want to take away that high from you. I just want to request you to add more tools in your arsenal and spend some time exploring new problems. Professors and researchers around the world, across departments are excited about the power of AI & ML techniques and very interesting work is being done in new directions.

  1. Health IoT
  2. Verification of autonomous vehicles
  3. Robotics for contagious diseases
  4. Weather prediction and chaotic systems
  5. Protein engineering
  6. Energy saving at data centers
  7. Theoretical understanding of deep neural networks

My goal here is to convince you to broaden your perspective about AI & ML problems and to urge you to try problems from different areas. After reading my semi-rant, you might still want to work on CV or NLP or the areas that I listed before. Please go ahead! You should be happy with whatever you’re working on.

Part 2: Seriously, what area should I work on?

Always remember that research is like long-term dating and not a marriage. You don’t marry a research area because it is okay to switch from one area to another. In fact, people have a wide array of research interests and excellent researchers are always trying something new. However, don’t mistake research to be a one-night stand, one has to give time to problems to make any meaningful contribution. It is okay to get enamored of multiple research problems for some time but you’ll have to calm down one day and dig deep into a few.

Many people including me are fearful of change. I have always been conservative and self-doubt takes over whenever I try something new. Taking too much risk is difficult but taking no risk is deadly. If you feel strongly about something, go for it. You’ll be amazed by how much you can accomplish. Easier said than done, right? I have always had close friends who have questioned my assumptions and forced me out of my shell. Peer pressure, if utilized correctly, actually improves you.

So, while choosing problems to work on, don’t be too scared and don’t be too wild. Be open and take a middle path.

What if I can’t find a professor whose interests are close to mine?

It is possible that you are interested in a particular problem that no one in the department or the institute works on. I’ve seen a few students who through their dedication and genius have been able to produce research even in such circumstances! I am not one of them but I can tell you what I did. A reasonable strategy in such circumstances is to work on problems that are close to what you’re interested in and you can find enough guidance in the institute. You could also move to areas that are close to your heart but are not necessarily your top choice.

I took the later approach. In my second year summer, I became interested in bioinformatics. I worked on a project in the Bio department with a nice group but the project had less computational and statistical components to it. Bioinformatics was my top choice then but I also liked the probability courses in the EE curriculum and I was aware of good professors in the EE department working on related problems. My third year intern helped me make the switch. I worked on the intersection of optimization and coding theory and I really enjoyed the internship. On coming back to insti, I started working on problems related to applied probability and have really enjoyed the last two years.

It is okay to switch areas but make sure that you give sufficient time to a project before moving on.

Finally, exploit your opportunities

I sincerely hope that you’ll be able to find something that interests you. If and when you do, give your best to the project. If you work on enough problems, you’ll definitely do something that is publishable and publications are very helpful when you apply to PhD programs. A clarification on publishing suggested by Chinmay Talegaonkar: excessive impulses to publish will necessarily have a negative impact in the long run, and most likely, in the short run as well. A good rule of thumb is to treat publications as important side products after solving an interesting problem. I wish you all the best for your endeavors and I hope that you’ll enjoy doing research.