Akshat Chaudhary (MT6)
Interned at IMT Atlantique, Rennes
Domain: ML in Cybersecurity: Application of Federated Learning
Motivation
I am a dual degree student, so I can intern for three years, particularly after the second and third years. But for single degree students, you can intern for only two. Generally, people in a dual degree program try a research internship after the second year. After the fourth year, they try to do the usual corporate internship, which they get through OCS. I have been particularly inclined towards ML since my first year. ML is a very research-intensive field with many opportunities for a research intern. I decided to opt for a foreign internship over SURA primarily because I wanted to explore a new country and culture. I was not very obsessed with research, but I felt the urge to examine research methodologically. If I had that chance and the opportunity to explore a new country with funding, I would always take it.
Application Process
I started applying in late December. Initially, I was emailing professors in the USA. Until mid-January, I had very few responses, all of which were negative. Then, I came to know about this particular professor in France. Luckily, he knew about IIT Delhi, and as an added incentive, my CGPA was decent. So, he decided to select me. Before going there, I didn't know much about Cybersecurity, but nobody expects you to have prior knowledge.
Preparation
The topic of my internship was Federated Learning, which is a pretty new research topic important in the field of privacy research that focuses on training ML models in multiple independent sessions. My supervisor, a PhD student, sent over some papers and suggested that I read them. He said that it was optional. Even if I didn't read them, it was fine. But, I would not advise going to the internship completely unprepared. Even though I didn't understand everything in one go, I still read all the papers. Having an abstract idea is very helpful. I already knew a bit of ML, which helped, too.
Intern Experience
It was quite a 'slow' internship in that deadlines were quite relaxed as opposed to corporate internships. It seemed boring sometimes, but consistency was the key, and the work was very rewarding. It helped that my internship was more focused on implementation rather than theory. So I could carry out my work on a purely experimental basis. My internship was exactly two months long. In the first month, my job was to implement some simple supervised and unsupervised learning models. The main part came in the second month when I applied these models for Federated Learning. My supervisor would usually give me two or three questions and the whole week to answer or implement them. I would have meetings every Friday. I mostly operated alone, and my supervisor's office was next to my lab if I had any doubts. During the week, I would read papers, mail their authors and think of different approaches to the problem. I also got the opportunity to present at a seminar, which was a new experience. Our end goal was to gather enough data to write a paper, but unfortunately, I couldn't finish it due to time constraints. So, I have decided to continue my internship online and hopefully publish the paper shortly. My workload was only five to six hours daily, with Saturdays and Sundays off. I had a lot of time to explore the place. Language was not an issue because a lot of people spoke English. Although, being a vegetarian, I found the food quite bland at times. The stipend was also quite decent. It was time well spent. Next year, I hope to do a research internship on the same topic of Federated Learning.
Advice to intern hopefuls
The best thing to do would be to start the mailing process as early as possible. Early November or even before that is an excellent time to start. Another advice would be to choose your topic wisely. Since research takes place in very niche areas and you'll be working on the subject for a month or two, you'll need ample motivation, which you would have only if you're interested. To understand your interests, I would suggest starting with an umbrella topic, for example, Physics, and then delving deeper to see what interests you. Visit the websites of the professors you're applying to and look at their research papers' summaries. You don't need to understand it completely; even the professor will not expect you to. If you do this with, let's say, 20-30 different professors, you'll find something that interests you. It would not hurt to do one or two self-projects in the area that you are interested in. I had also done a self-project in ML. You can add this to your resume, along with your course assignments.
If you're a dual degree student, being one myself, I would recommend taking up a research internship this year because you'll probably not be applying for the internship through OCS come the summer. That time could be constructively spent in a research internship. I recommend being cautious for the four-year students because the internship may clash with your preparation for the OCS internship. Set your priorities straight beforehand.
Goals for the future
Being a business-minded person, I plan to do an MBA in the long run. For next year, I am considering applying for a research internship at MITACS, a good research program in Canada. After my fourth year, I will probably try to do a corporate internship through OCS.