Research Interests
I am interested in performance and power optimization of DL/ML workloads for custom accelerators and GPUs. I am also interested in leveraging compiler technology for enhancing security/reliability of applications.
During PhD, I worked on GPU compilers trying to develop new techniques for predicting and improving reliability of applications. My interest in compilers began while I was pursuing co-op at AMD. There, I worked as a Performance Compiler Engineering Intern in AMD's shader compiler team at Boxborough, MA in 2013. I worked on various optimizations for graphic and compute shaders. In 2015, I had the opportunity to work with AMD's Reliability Research team to leverage compilers for improving resilience of GPU applications.
In the past, I have also worked on other projects such as big data, power and performance evaluation on GPUs, GPU benchmarking and runtime systems. I enjoy doing inter-disciplinary research by collaborating with other group members on various projects. This not only helps me diversify my research, but also gives me a wider perspective during problem-solving and a sense of appreciation towards other domains of research.