Happy New Year everyone!

I hope everyone had a great year! 😀 Today’s blog is about some thoughts I had regarding my internship last year at ARM. I spent two months from August to October,  interning at ARM research in Cambridge. If you are a PhD student like myself or are thinking of doing a PhD you might be interested in knowing about such opportunities whilst you’re studying. I think some of the questions that may be boggling your mind is how to apply for one? When is it advisable to do it? (Especially,  because we only have 3 years to complete a PhD program in the UK). What should the internship be about? How long should the internship be?

But out of all these questions, I think the most important one is whether taking the time to do an internship is beneficial towards your actual thesis. Most people say that a PhD is a long commitment but after having spent two years, if anything, I feel I need more time!

This was one of my main concerns when I decided to do the internship. However, after discussions with my supervisors and agreeing on a topic area that would be part of my final thesis, I was given the thumbs-up to go ahead.

It was when I visited ARM in February last year, for two weeks that I began scouting for projects that might interest me, as well as align with my topic area. Another important aspect of my internship was finding someone who would be happy to work with me and I found out that the best way to do this was by networking and developing connections, especially at conferences. So, I spoke to someone who knew someone, and so on and so forth. 😀

There are of course other channels by which you can find such opportunities. E.g. HiPEAC offers internships.

I decided to do my internship whilst transitioning from my second to third year. This is because I felt at the end of my second year, I would have an enhanced understanding of my research and would have started making progress in that particular direction. However, at the end of the first year, I was aware of the general area I would be working in, for example I knew my research would be at the intersection of applied AI and systems research. I had covered  a substantial literature review pertinent to the field, right from accelerating deep learning from the software level on existing machines to building custom hardware accelerators.

So, I think choosing a suitable time to place your internship would allow you to best exploit the opportunity. Now, I have narrowed down my research area to deep learning on low-powered devices, like mobile phones with the aim of managing trade-offs in performance and energy (or battery-life). Even within this area there are myriad sub-fields I could be working on, for example:

Do I design better algorithms?

How do I optimize existing software implementations and find ways to approximate certain computations?

Do I tune the characteristics of the hardware for the application?

Finally, I had to decide on an apt question and just go for it!

Doing an internship during the third year might be a bit stressful as it is the time that your stress-levels are off the charts trying to do experiments, writing up your thesis and handling everything else in between. Now that I am writing about it, I am starting to feel all anxious!

As for my internship, I chose a topic that closely matched my research objective wherein I worked on exploiting task-based parallelism in deep learning algorithms like Convolutional Neural Networks and understanding the manner in which to execute these task-graphs efficiently.

I interned at ARM for 8 weeks, as this was the maximum number of weeks I was allowed to do an internship under student visa rules. Simple as that! I think internships for Home students otherwise can be of minimum 3 months stretching up to 6 months. However, as I have noted earlier, you should make sure the work done during the internship is preferably something that you can add to your final thesis!

Apart from that, during my internship I also attended IISWC and the ARM Research Summit to present my work. The summit really helped me learn more about ARM and the research opportunities. Overall, I had a very enjoyable time at ARM. It was very much similar to my experience during my two weeks in February.

Cheers,

Crefeda