Applied research in deep learning - Duplex: task oriented dialogue system to accomplish real-world tasks over the phone, including making reservations through Assistant and providing accurate business information on Maps. - Vehicle Damage Solution: computer vision based system for nearly instant vehicle damage estimates from images. - Recsim: the first open-sourced configurable platform to facilitate the study of RL algorithms in recommender systems. https://blog.google/inside-google/company-announcements/covid-19-how-were-continuing-to-help https://ai.googleblog.com/2019/11/recsim-configurable-simulation-platform.html https://www.repairerdrivennews.com/2019/09/27/usaa-works-with-google-cloud-mitchell-on-ai-photo-estimating/ Duplex helps with COVID-19 RecSim: A Configurable Simulation Platform for Recommender Systems USAA works with Google Cloud, Mitchell on AI photo estimating

We are enabling a new generation of intelligent software built by the masses and powered by an open shared platform. https://www.scaledinference.com/

Developed the high frequency trading strategies for foreign exchange (FX) markets.

Developed a spectral clustering algorithm for clustering users based on the pairwise comparisons of items they provide. Implemented fast algorithms for solving trace norm regularized convex optimization problems.

Developed a Hidden Markov Model (HMM) for WiFi indoor positioning. Implemented a generic belief propagation algorithm to infer mobile device trajectories. Built a system to evaluate the visible light communication (VLC) channel condition in various room configurations using ray tracing techniques .