Research

I have worked on a diverse set of research topics. Please follow the links below for more details

Current Research Projects

  • Accelerating Reinforcement Learning on Reconfigurable Platforms (FPGA): The objective of the project is to develop a unified Application Specific Processor (ASP) for Deep Reinforcement Learning (DRL) applications targeting FPGA. This project is funded by the U.S. National Science Foundation under award number 2009057.
  • Accelerating Homomorphic Encryption based Deep Neural Network on Reconfigurable Platforms (FPGA): The objective of this project is to develop an end-to-end implementation of low latency Homomorphic Encryption based DNN inference on FPGA. Homomorphic Encryption allows computations in encrypted space. Thus, this project will allow privacy sensitive applications to utilize the excellent representation power of DNNs leveraging the vast and economical computational capability provided by public cloud platforms in a trusted and secure manner.
  • Reinforcement Learning Based Decision Making for Smart Power Grid Operations: The objective of the project is to develop Reinforcement Learning algorithms for Building HVAC scheduling that reduce the cost of operations while ensuring safety and comfort constraints are not violated.
  • Data-driven Spatio-Temporal Modeling of Smart Power Grids: The objective of the project is to develop Spatio-temporal Graph Convolutional Neural Network based models for smart power grids. These models, which capture both spatial and temporal correlations, can enable accurate short term forecasting and missing data imputations.
  • Accelerating Graph Analytics on Emerging Cloud Architectures with CPU+FPGA Heterogeneous Nodes: The objective of the project is to develop high throughput implementations of Graph Analytics applications targeting cloud architectures with CPU-FPGA datacenter nodes. This project is funded by the U.S. National Science Foundation under award number 1911229.

Previous Research Projects