By Sriram Rajamani, Distinguished Scientist and Managing Director, Microsoft Research India at Microsoft
Manik Varma, researcher in the Microsoft Research (MSR) India lab was awarded the prestigious Shanthi Swarup Bhatnagar (SSB) prize for 2019. The SSB prize is a coveted prize for multidisciplinary science and engineering in India. He was also elected to the Indian National Academy of Engineering the very next day. These honors add to a long list of recognitions that researchers from Microsoft Research (MSR) India have received over time, including:
- SIGCHI societal impact award in 2017: Indrani Medhi-Thies, for work on interfaces for low literate users
- Shanti Swarup Bhatnagar prize in 2016: Venkat Padmanabhan, for pioneering work on indoor localization, smartphone based sensing, and mobile communication.
- MacArthur grant in 2016: Bill Thies, for advancing the socioeconomic well-being of low-income communities in the developing world through innovative communication and digital technologies that respond to real-world constraints.
- Polya prize in 2014: Nikhil Srivastava (who is now a faculty member at UC Berkeley) and his collaborators for resolving a long-standing open problem called the Kadison-Singer conjecture
- Knuth prize in 2011: Ravi Kannan for with many powerful new algorithmic techniques, which have enhanced our understanding of computational complexity, discrete mathematics, geometry, and operations research.
- Computer Aided Verification award in 2011: Sriram Rajamani (with Tom Ball), for contributions to Software Model Checking
These, and the other honors our researchers have won over the years are indeed a source of great pride for us. At the same time, we recognize that almost none of this would have been possible without the great collaborations we have with our academic partners across the globe. MSR, since inception, has always recognized the importance of collaborations with academic partners and other researchers, not only for advancing the state-of-the-art, but also for their incisive feedback on our research quality and directions. In fact, MSR India has a Technical Advisory board that features some of the world’s leading academics to advise us on precisely this- the quality and direction of our research.
Let me use Varma’s (our most recent honoree) work as an example. He got his PhD from Oxford University in Computer Science with specialization in computer vision in 2004, did a brief postdoctoral stint at UC Berkeley, and joined our lab soon thereafter. His initial work at MSR India continued to be on computer vision–he worked on detecting objects in images and won competitions in the area. But he soon switched over to Machine Learning and the bulk of his work in the past decade has been in Machine Learning. He has done work on local deep kernel learning, which received both academic recognition and found practical utility; a virus classifier based on this work is widely used.
However, his most impressive and impactful of contribution over the past decade is in Extreme Classification- a sub-discipline of Machine Learning which Manik helped originate. Instead of traditional classification algorithms, which classify objects with a small number of labels (categories), Varma has been building classifiers which classify objects with a large space of labels- think millions of labels! This formulation with a large space of labels makes sense for several scenarios such as phrase selection (for advertiser bidding, for example) in online advertisements, and tagging Wikipedia articles with tags, where the space of phrases or tags is very large. The initial algorithms used an ensemble of trees with lots of tricks to help scale while maintaining accuracy. More recent algorithms are based on embeddings generated by Deep Neural Networks.
He started this area and, together with colleagues at Microsoft and students at IIT Delhi, wrote the first papers on this topic. Over the past decade, we have seen papers by Varma and his collaborators on this topic routinely appearing in conferences such as ICML, KDD, NIPS, SIGIR, WSDM and WWW, with some of the papers winning best paper awards and receiving considerable attention and recognition from the community. More than ten workshops have been organized in the last seven years in various conferences on this topic, and he has been the keynote speaker in most of these workshops.
He also collected benchmarks and created an “Extreme Classification Repository”, which has been viewed and downloaded thousands of times every month and is now used as a standard to measure algorithmic progress in this area. In addition, he has been a key contributor in building ML algorithms that run on very small devices, with very low memory and power footprint.
Varma is an ambassador for Microsoft Research and the values and principles our organization stands for. Much of the inspiration for his work comes from the real-world problems faced in industrial settings. He is also an adjunct faculty member at IIT Delhi, and much of this work was done in collaboration with PhD students at IIT Delhi. Thus, Varma’s award is a celebration of not only his work, but the collaborations he has had with academia as well as colleagues at Microsoft.
We live in interesting times where industry and academia have a great opportunity to work together to make scientific progress. Some problems that arise in industrial practice are open ended and need many years of sustained research to make progress in solving them. Exposure to such problems greatly benefits students and faculty in academia and hence such collaborations are mutually beneficial.
MSR is committed to continuing collaborations with our academic colleagues, to expose and formulate difficult problems that arise in industrial settingsand bringing to bear the collective wisdom of the academic community and advancing the state of the art to solve such problems.
For more information on MSR India, employment opportunities and collaborations, visit our website.
