Watch Out!
Student Media
Body of IITR
About Guide Get Involved

Summer 2018

Summer Diaries: University of Michigan

August 25, 2018
- Ekdeep Singh Lubana

I spent my summers working as a visiting scholar (the appropriate term that should be preferred over the standard lingo—Research intern) in the Embedded Systems group at University of Michigan. My project encompassed working on the brink of software- and hardware-design in order to reduce energy consumption of a camera.

With a GPA of 7.05, I certainly did not have high expectations with any of the conventional means of bagging an intern. However, a decent research background that had resulted in significant output kept me hopeful. Thus, I started mailing professors around the last week of August.

Luckily, I did not have to squander a lot of time mailing and faculty-hunting. While my first two applications went dark (I still haven’t received a reply from them quietly sobs in denial), the third application got converted into a teleconferencing interview. As I mentioned before, I believe, and my advisor has affirmed this multiple times, that relevant research experience in the field of imaging systems is what bagged me the interview. After that, I spent a week brushing up my computer architecture and digital design basics. The interview went fine and my advisor, based on my previous experience, told me to come up with ideas on a biological principle named “fovea”.

Off I went and figured how energy consumption in machine vision systems can be significantly optimized if a fovea-like sampling routine—i.e., multi-resolution sampling of analog image signal—is used. Essentially, this results in tight integration of the software algorithms with hardware routines that exist, but are not exploited. This proposition was well received and I started working in remote capacity. Here, I would like to emphasize that most faculties are apprehensive of research interns because of limited time windows. My explorations in remote capacity helped me convince my advisor to have me at U. Mich. as a winter-cum-summer intern. The “wintern” work resulted in 75% energy reduction of conventional machine vision systems and a publication on the same was accepted to IEEE Transactions on Computer-Aided Design. The paper will be presented at CODES+ISSS (International conference on co-design of hardware/software and system synthesis), 2018—a part of Embedded Systems Week, that is set to be held in Italy (Yay!).

pic2

Work Culture

Located in Ann Arbor, University of Michigan is in a relatively remote region that has more folk-culture vibe than a trendy city like New York. That suited me really well. Further, there were no time restrictions and we weren’t even expected to show up—at all. Thus, I reckon I spent my entire time meandering in the library, that had amazing rooms for private study, or CS lounge (or as it is called—the ”foo bar”) which had an amazing ambience. I love this minimalist nature of computer science that renders extravagant resources unneeded. Departments and library were open 24x7 and were connected to my home via the Michigan M-bus route. Thus, I would reach campus by 8 AM and wander around till 12-2 AM, mostly.

We had research group meetings every fourth Monday and reading group discussions every Thursday, wherein the leading participant would propose a paper on the field of Embedded Machine Learning and the same would be discussed in the meeting. I met my advisor twice a week. While meeting your advisor twice a week is horrendously exhausting, it helps accelerate the overall project. I expected myself to show some progress every meeting and that helped produce significant output. I completed a project in the winter, had a paper accepted, and spent my summers working on really interesting aspects of Embedded Machine Learning and Neuroscience, wherein the focus was to come up with ideas inspired by the latter and make the former efficient. This has led to a novel data compression algorithm based on compressive sensing.

pic3

Ann Arbor

Not a lot happens in Ann Arbor. Alas, being a foodie, I loved the minuscule vegetarian serving places in Downtown. Being a vegetarian is a bummer, take my word for it. I would recommend trying the Indian eateries like Madras Masala to satiate your desires for Indian food.

The Ann Arbor summer festival and art festival were fun and a must visit. There’s a record store that sells vinyl records and CDs of the oldest artists you can think of. I visited the place quite often. Bookstores in Ann Arbor are really interesting, for if you visit on first or third Saturdays, you can find Bruce Conforth playing folk music (Heads up—Bruce Conforth is the first curator of rock and roll hall of fame and has worked with the likes of B. B. King, U2, The Kinks, etc.).

pic4

Takeaways

In a nutshell, I would recommend to try your hands at research internships if you either have relevant experience or high GPA to bag one via conventional programs. I believe the former is more sensible, though. That said, avoid spamming professors.

I could recommend two methodologies for mailing: (a) Pick a university using QS or some other organization’s ranking system. Narrow down on faculties with relevant interests. Visit their Google Scholar or dblp profiles and go through the recent research work. If the interests align, draft a mail explaining how your interests match their research directions. Herein, boast. (b) If you are deep into your field, reading papers is a routine habit. Off of the recent year conferences, if you find interesting research papers, you can explore the authors’ previous work and draft a mail, thereafter. This methodology helped me bag another internship for Winter, 2019.