About me

I’m a Research Fellow at Microsoft Research India. I’m passionate about machine learning, computer systems and robotics. In the past, I’ve dabbled in a bit of reinforcement learning and natural language generation systems.

Checkout BLAS-on-flash - a library to describe and execute computation graphs for data residing on disk. Our work has been accepted to NSDI ‘19 and SysML ‘18.

I’m currently working on two problems involving systems and machine learning.

  • Ensemble-distillation procedures to enable training of deep relevance models on web-scale data with over 2 Billion training points with Dr. Harsha Simhadri, Dr. Praneeth Netrapalli
  • Developing a new algorithm to efficiently serve k-Approximate Nearest Neighour Search (k-ANNS) queries on over 10B vectors on a single multi-core system utilizing next-gen flash drives with Dr. Harsha Simhadri, Dr. Ravishankar Krishnaswamy, and Dr. Amit Deshpande

Research Interests

  • Machine Learning - algorithms for large-scale machine learning, deep learning, reinforcement learning
  • Computer Systems - parallel processing, distributed processing, large-scale machine learning, distributed machine learning systems, next-gen machine learning systems
  • Robotics - multi-agent robotics, single-agent robotics, reinforcement learning for single-agent and multi-agent robotics

Contact

Email: j.s.suhas AT live.in, t-sujs AT microsoft.com

LinkedIn: Suhas Jayaram Subramanya