I have a wide interest in Machine Learning spanning from theory to practice to system designs.

My current main research pursuit is reinforcement learning (RL) with focus on model-based reinforcement learning, uncertainty quantification, and planning/decision making under uncertainty; and real-world optimal control applications, including controlled environment agriculture (CEA) optimization, energy optimization, etc.

Previously, I’ve also worked on:

  • Convolutional neural nets and computer vision (object detection, image segmentation, image classification, etc.)
  • Image captioning and multimodal modeling
  • Optimization algorithms
  • DL/ML platform

Below are some of the projects in which I’ve played a leading role:




With the world’s population growing rapidly, it will become more difficult to produce food quickly and sustainably in the amounts needed. By 2050, we’ll need to increase our food production by at least 50 percent. Greenhouses and indoor farming provide a solution to produce fruits and vegetables faster, more safely, and with much fewer land and water requirements. However, it is difficult to find enough skilled labor for greenhouse production. Automation is an answer, but…

Model-based Reinforcement Learning for Control Problems

Established: August 1, 2017

This research project aims at developing a new class of Reinforcement Learning (RL) algorithms that are sample efficient, off policy, and transferable. We seek to demonstrate these new algorithms in real-world operational optimal control applications such as Indoor Farm Control     Data Center Energy Consumption Optimization News Microsoft Asia News Center covers our research in indoor vertical farming

Vision and Language Intelligence

Established: June 28, 2017

This project aims at driving disruptive advances in vision and language intelligence. We believe future breakthroughs in multimodal intelligence will empower smart communications between humans and the world and enable next-generation scenarios such as a universal chatbot and intelligent augmented reality. To these ends, we are focusing on understanding, reasoning, and generation across language and vision, and creation of intelligent services, including vision-to-text captioning, text-to-vision generation, and question answering/dialog about images and videos.

Time Series Analysis

In this project, we aim to develop a time series analysis framework using modern machine learning techniques. The project focuses on solving the two fundamental tasks in time series analysis. 1. Multivariate Demand Forecasting With Uncertainty Estimate Our approach is based on Bayesian LSTM. Besides direct forecasting applications, our research also studies time series analysis in the context of non-markovian reinforcement learning. The most common framework for RL relies heavily on the markovian-observation assumption, that…

FarmBeats: AI & IoT for Agriculture

Established: May 14, 2015

Several studies have demonstrated the need to significantly increase the world’s food production by 2050. However, there is limited amount of additional arable land, and water levels have also been receding. Although technology could help the farmer, its adoption is limited because the farms usually do not have power, or Internet connectivity, and the farmers are typically not technology savvy. We are working towards an end-to-end approach, from sensors to the cloud, to solve the…


Get in Touch