I am a principal researcher at Microsoft Research Cambridge. Before joining Microsoft Research, I was at Disney Research Pittsburgh located at Carnegie Mellon University. I have received my PhD from the Department of Robotics, Perception and Learning (RPL/ former CVAP), KTH Royal Institute of Technology. My general interests are Machine Learning and Computer Vision. I am currently working on Representation Learning and Approximate Inference.
Efficient element-wise information acquisition
Modern AI works well with large amounts of data, but can it work when data are scarce and expensive? Our work adopts Bayesian deep learning and Bayesian experimental design to address this setting. Data acquisition in many real-world applications is difficult – challenges vary from increased time and costs to potentially posing privacy risks. In this project, we investigate how to utilize AI algorithms best to aid decision making while simultaneously minimize data requirements and, therefore, cost. In Microsoft, we developed novel frameworks for efficient decision making using Bayesian experimental design combined with novel deep generative models to solve it.