How to Get a Summer Research Internship

11 minute read

Recently I spent some time looking for summer research internships and thought that it might be useful to share some of my experiences from the process. I will talk about how to choose where to apply, how to apply, and what to do if you get rejected. So, if you’re thinking of applying for a summer research internship but don’t know how to go about it, or if you had a bad experience last time you applied, this post is for you!

First, I think it will be useful if I tell you a bit about my circumstances which might help to give you some context. A few weeks ago I have finished my 2nd year of physics undergraduate at Oxford University and was looking for a summer research internship related to machine learning (which is not really related to physics but that’s what I’m interested in). So, after having finished 2nd year I have spent almost 2 full weeks emailing and visiting people to see if they would be interested in supervising me over the summer and here are some of my observations. (In the end I ended up doing an intern in deep learning at the Visual Geometry Group at Oxford.)

Choosing where to apply

The first thing you have to think about is what area you’d like to do your summer research in. This should be something you find really interesting or something which you want to learn more about. Since you’ll probably spend the next 2 to 3 months doing this make sure it’s something you would enjoy. For me that area was machine learning (or more specifically deep learning) because I find it really interesting and want to learn more about it since I’m thinking of doing a PhD in it. Note that it doesn’t have to be related to your major (as in my case), although if it is it probably helps.

After you’ve decided which area you want to do your internship in, you need to decide where to apply. So, in the case of a research internship, you need to find out which research groups at which universities are focusing on the area that you’re interested in and explore those. The way I did this is I searched for something like “oxford deep learning” and went through the links to get an idea what groups and people does it have that are related to deep learning. I did a similar process for other universities, for example searching for something on the lines “cambridge machine learning group”. As an example, let’s say I found an interesting group at Cambridge focusing on Natural Language Processing: https://www.cl.cam.ac.uk/research/nl/people/.

After finding an interesting group I went to the group’s website and looked at what people are working in the group. In my case I went through the group’s hierarchy from top to bottom, i.e. starting from professors, then post-docs and then PhD students and looked at each person’s summary of research interests. If I found their research interests interesting, I went to their website and explored their research further by looking at their stated research interests there and by going through the titles of their publications to get an idea of which areas they’re interested in, sometimes also checking their CV. If I found that person’s research interests matched mine I have noted the information about the person down in a table that I kept to keep track of people that might become potential supervisors. For example, in the case of the above mentioned Cambridge NLIP group I first went through the Staff section, then Fellows section and then Post-docs section and noted down Prof Ted Briscoe, Prof Simone Teufel and Dr Marek Rei together with their research interests, link to their website in case I needed to do some further research on them, and their room number (I was planning on visiting some of them in person). I did this for a few people from each group and at the end had a table which looked like the following, with the most suitable people highlighted in red (this table is just for the Cambridge groups but I had similar tables for Oxford and other universities).

Cambridge AI related groups

How to apply

After I had found enough people with interesting research interests I needed to let them know that I’d like to do summer research with them. The best way to do this is probably to email them and so I started emailing the most interesting people first, and while I was waiting for their replies I looked for more people who might be doing interesting research and kept adding them to tables.

So, how to go about writing such emails? The email is very important since this will probably be the first (and maybe the only) place where the researchers will get to know you and your experience. So, it’s important to make sure you give them the right information in a concise way so as to tell them that you’re qualified for the position without wasting too much of their time. The way I structured these emails is roughy as:

  1. short introduction to yourself — who are you, what do you study, what research areas are you interested in
  2. why are you emailing them — which of their research areas do you find interesting, what do you want from them (summer research), possible setup for a meeting
  3. your experience — a longer paragraph listing your relevant experience which may include previous internships, online courses/tutorials or links to your projects on GitHub
  4. action they need to do — make clear what you want from them again in a short sentence and sign off

It is important to keep the email as short as possible because professors and postdocs are usually very busy and have very little time to reply to undergraduates, so I tried keeping points 1., 2. and 4. very short with a slightly longer point 3. which they can read at their leisure (although still relatively short). I learned my lesson after last summer’s applications where I realised that I didn’t get many responses to longer emails despite the fact that they contained more information, most likely because the people didn’t have time to read them. On the other hand you want to give enough information about your experience to tell them you’ll be able to handle the work you’ll be doing and express your interest in the area.

As an example, here’s one of my emails to one researcher:

Dear ____,

I’m Marian and I just finished my 2nd year of undergraduate physics at Oxford. I am interested in doing summer research related to NLP and ML and found your research interests interesting, especially text understanding and summarization. Would you be free to meet in the Computer Laboratory or elsewhere in Cambridge to discuss possible summer research? I am in Cambridge today and tomorrow but could also come back next week.

My experience: although I study physics I have wide experience with computing, having worked on many CS-related projects (see my list of projects or GitHub). Last summer I have worked at Imperial College London’s Data Science Institute where I have used machine learning in a project to detect fake news tweets based on their metadata (where I worked with Python, R, Scikit-Learn ML library, etc). I have also watched an NLP tutorial series which uses the NLTK library with Python to explain the basics of NLP together with simple sentiment analysis. I am also very excited to learn anything new which would be useful for the project.

Please let me know if you would be interested to meet. Thank you.

Best regards,

Marian

After you send your first email the further ones are going to be much easier since you can just modify some of the information to match the particulars of a different researcher but can still keep some information the same.

After they reply to your email (if they do indeed reply) you should arrange a meeting in person if possible. This part is also important because you can get to know the researcher, what kind of research they’re doing in more detail, and you can have a discussion about possible projects that you’ll be working on and choose a right one for you. In case they suggest a project you’re not that interested in you should tell them so and they can think of some other project which will be more suitable since they’re usually interested in different areas of a particular field. Hopefully by the end of the meeting you should have a project that both you and your supervisor are happy with.

Success rate

This is the ideal case — you send a few emails, arrange some meeting and finally get an interesting project. However, things might not work out this way: the person you emailed might reply that they’re not looking for interns at the moment or that they’re fully booked for the summer or they might not even reply at all. These are actually what happens a lot (to me, anyway) and it shouldn’t make you disappointed as it’s actually quite understandable from their perspective: imagine you’re a professor supervising many post-docs, perhaps having a few trips planned over the summer and an undergrad with almost no experience emails you about doing research with them — what would you do? So, although these cases are common there are likely to be at least a few people who do reply and although the proportion is quite small it doesn’t matter since you’ll only be doing research with one of them.

As an interesting experiment I have compiled a statistic of how people responded to my emails, categorised as no reply (if they didn’t reply), no opportunities (if their institution doesn’t have any relevant research opportunities for me), full (if they’re fully booked for the summer) or successfully referred (if they referred me to another person, usually a PhD student, and that was successful). Note that none of the replies was actually directly successful with the professor I emailed but they instead referred me to one of their students and that was successful. I also got a slightly different response when meeting the people in person. Also, you should note that this data is not at all statistically significant to draw any conclusions from as I only have 12 samples, but it’s useful to show that many of my emails were unsuccessful in some way.

Table of replies to my emails about summer research

And here is the data in the “email reply” column of the table visualised:

Chart showing type of email reply

If I remember correctly when I was applying last year for summer internships (for companies as well, not just for research) the statistics were even worse — out of around 15 applications that I submitted I was only successful in around 3 (but back then I have just finished my first year and I was also applying to the most competitive companies). In any case, all of this is just to show you shouldn’t care too much about getting rejected since it happens a lot 🙂

Some other thoughts

Should you apply to better groups that are harder to get into or lesser-known groups that are easier to get into? It might be the case that it’s harder to get into the best groups for a summer internship than to some lesser-known groups, so why not just apply to the lesser-known ones? There’s nothing wrong with applying to lesser-known groups if you think you’ll enjoy your time there. However, in my opinion there are some benefits of going for the top groups, such as: you’re probably more likely to be working on more interesting research, you’ll be with people who are top in their field and be able to learn from them, and it’s also better CV-wise. Also, although there’s a lower chance of getting accepted to a top group, you need to find just one person to work with, which you probably will if you send enough emails (and I find the extra time spent on this is worth it).

Visiting in person or just emailing. It seems like when most people are applying for a summer internship they either apply through some official listing of projects or just email the researcher. However, I feel like most people don’t actually meet with the researchers in person when they want to get a research intern (but probably only after they get accepted). From my experience this is one of the best ways to increase your chances of getting accepted, or, if nothing else, at least make a potentially useful connection and learn more about their research. The fact that most people (I assume) don’t visit the professors in person to talk about research opportunities gives you a big advantage if you do, since it shows that you’re motivated and interested in the position enough to spend your time talking to them instead of doing something else, and you might also get a position which is not officially listed online. So, from my experience I would highly recommend this.

Working with a professor vs post-doc vs PhD student. Although initially I thought it’s better to work with a professor than with a post-doc than with a PhD student, I’m not sure this is actually the best way to do it. Sure, the higher in hierarchy you go the more experience will the person have with research but also the less time they will have to supervise you. So, you want to choose someone who is both very knowledgeable about your area of interest so they can give you a meaningful project and provide useful help, but also someone who will actually have the time to supervise you. So, personally, I think that although most professors are very knowledgeable they are very busy and although PhD students have more time they are also less knowledgeable (obviously this is a wide generalisation). So, in my opinion it may probably be best to work with a post-doc for to get a good knowledge-time balance, although depending on your and their experience working with PhD students or professors might also be a good idea in certain cases. Ultimately, it depends on each individual.

Conclusion

So, in summary, do your homework and find researchers who are interested in your areas of interest, then send them an informative but to-the-point email and meet them in person if you have the opportunity, and finally don’t get discouraged if they don’t reply. Good luck in finding your next summer internship!

Leave a comment