A peek into the Industrial PhD Program

Hi everyone, I’m Samson, a PhD student in Salesforce’s Industrial PhD Program jointly offered with the National University of Singapore (NUS) and co-sponsored by Singapore’s Economic Development Board. In this post, I share my experience so far, both to mark the end of an exciting first year and hopefully give aspiring students an idea of what it’s like to be a Salesforce IPP trainee. For those who have not heard about it, the Industrial PhD Program is an initiative by Singapore’s Economic Development Board to support joint partnerships between companies and top Singapore universities to nurture a pool of AI talents for future R&D roles. After setting up Salesforce Research Asia as its first APAC AI research centre in Singapore, Salesforce partnered with three top Singapore universities and joined the initiative last year with me as part of the inaugural batch!

Now, you might be thinking: why do a PhD when there’s a plethora of high paying software engineering jobs out there? You’d be right, a PhD is definitely not the right choice for most people due to personal interests and the huge opportunity cost. Even though I’ve always enjoyed reading, tinkering, and writing, spending another 4-6 years pursuing a PhD would not have been financially viable for me under normal circumstances.

However, the IPP’s competitive stipend and the opportunity to learn from top AI and Natural Language Processing researchers here made it a no brainer. I was also attracted to Salesforce Research due to its strength in natural language processing, focus on the ethical use of technology and using AI for social good, and the possibility of collaborating with researchers from its U.S. lab. Other significant draws for me were the ability to work on a fundamental research area of my interest while being exposed to how machine learning is used in production, the computational resources available, and its welcoming culture.

The solid mentorship and ever-helpful coworkers at Salesforce Research Asia enabled me to publish a full length conference paper at a top NLP conference in my first semester even though I had never written one before embarking on my PhD. I could always count on my mentor, Assistant Professor Shafiq Joty, to offer guidance in areas I was unsure about while giving me the freedom to pursue the topics I was interested in. After some reading and deliberation, I decided to focus on making NLP (and AI) systems more inclusive and robust.

My friendly fellow researchers (from both Singapore and Palo Alto) were always ready to explain key ideas and remove any confusion I had. When the conference submission deadlines were approaching, I could also count on them to help with internal peer reviews. In one case, we even ended up collaborating on a future paper due to our overlapping research interests! Without their help, it would have been nigh impossible for me to publish a paper in my first year.

As part of the IPP, I divide my time between NUS (mainly for classes) and Salesforce Research. A key feature of the joint program is co-supervision by a professor from NUS' School of Computing. To get the most out of this program, it is crucial to choose an academic advisor who complements your Salesforce mentor. With Associate Professor Min-Yen Kan as my advisor, I have the best of both worlds: his emphasis on micro-analysis and Shafiq's machine learning lens.

Although it was definitely a challenge at times to balance both coursework and making progress on research, having access to sufficient compute resources (8 top end GPUs all to myself!) ensured that my time was spent running experiments instead of waiting for GPUs to be available. That said, all work and no play makes for a dull experience, hence in between bouts of hard work were also monthly team lunches and occasional team outings for us to take breaks from work to get to know each other better!

I hope this post has given you a better idea of what it’s like to do a PhD with Salesforce Research! In a later post, I will write about my recently published work at ACL 2020, a top conference for natural language processing research.