Staring into my laptop at the makeshift workspace I have salvaged at home; this was definitely not how I had expected my internship to be, but, to use the hackneyed phrase, in these unprecedented times this was the best that could be done. Hi, I am Anshul Nasery, a third-year undergrad in the Computer Science Department. I am a research intern at the Big Data Experience Lab at Adobe, and this is the story of my internship, so far.
Bitten by the ML Bug
Since my first year, I have been interested in machine learning and have been inclined towards research in this field. Having done univ interns in my second-year summer and third year winters with groups working in this area had reinforced this penchant, but I also wanted to gain experience of industrial research, to find out how it would be different from an academic setting, and to decide if I would want to pursue either of these paths further. Hence I decided to apply for industrial research internships in my third year, especially those which had a significant machine learning component. Among the companies coming in through the PT Cell, Adobe Research was one of the profiles which fit the bill perfectly. The pre-internship talk for the position highlighted some of the past projects that had been completed by interns. These had not only led to academic papers and patents, but were also a part of products and were being used by millions. To be able to work on research problems which would have a real-world impact seemed very exciting to me, and I applied to the position.
The selection process
Adobe usually has two streams for intake of interns. They offer internships directly to the top rankers in the Computer Science and Electrical engineering departments, and they also have interviews for others. The first stage of the latter process was shortlisting based on the resume, which was followed by a telephonic interview. The interview consisted of some basic data structure questions, some logical puzzles as well as some discussion on past projects on my resume. It was a relatively simple interview, and most of the shortlisted candidates were selected for the internship.
The project that I am working on is a natural language understanding problem of building a question answering system on multi-modal documents. We are working in a team of four along with two mentors. Due to uncertainties about the exam schedule etc. of various institutes, the company tried to allocate teams in such a way that all members of the team are from the same college. The teams were then allotted projects and mentors in a (pseudo) random manner. While we were given an initial high-level overview of the problem topic, it was up to us to decide what exactly was the problem statement we wanted to work on. The first two weeks were spent in coming up with this problem statement. This was accompanied by an exploration of what sort of problems had been solved in this space earlier, determining the feasibility of our proposed project, and most importantly, how would our project benefit Adobe and its customers. The last point is one which is seldom talked about in academia but had to be emphasized in our problem proposal. Once that was done, we set out to build a solution to our proposed problem, and workdays now typically consist of reading research papers, discussing ideas, and coding them up to experiment.
The work – from home
Working from home is obviously a new experience both for us and the company.
The major problem I faced in this scenario is that communication is difficult, both formally and informally. While there are some arrangements set up in this regard in the form of daily sync up video conferences among the team, having small chats to discuss ideas or to collaborate on code often becomes clunky due to the overhead of scheduling and setting up video calls. In a research environment, collaboration through regular discussions is essential for better exchange and development of ideas. While that is theoretically possible in a remote setup, it is often less effective due to logistical factors.
Further, interactions with other interns from other teams and institutes, as well as other employees of the organisation have been almost non-existent. Such interactions were some of the most valuable experiences from my previous interns, and working from home did put a dampener on them. However, Adobe did make efforts to make sure that we do not miss out on such networking opportunities (more on that later).
On the plus side, working from home gives quite a bit of flexibility in terms of work schedule. Since life in the insti has messed up our circadian rhythm, most of the interns (and mentors too :P) end up working at night, arguably leading to better overall productivity.
Adobe as a workplace
The overall experience of working at Adobe has been great. They were among the first companies to assure us that they would be going ahead with virtual internships despite the overheads it would result in, and the admin has left no stone unturned in ensuring that we have the best internship experience possible. The work culture is pretty conducive to research, with mentors encouraging novel ideas, weekly research group discussions on projects around Adobe, and regular feedback from upper management about our work. To alleviate the social effects of a remote intern, the company also regularly organises fun virtual events like quizzes and tambola, as well as intern meetups to promote interaction among various teams. They have also conducted multiple workshops focussing on skills required in the corporate world to prepare us for a future in the industry.
While the conditions for the internship have not been ideal, this intern has taught me some valuable skills. Adapting to a new working environment, presenting research problems in a way that they are relevant to an end-user, and working in a team which is spread across the country have all been a part of this learning experience for me. While the technical learnings from the project that I am working on have been satisfying on their own, the other experiences are the ones that I feel would be more relevant going forward, and it is these learnings that make an internship useful.