04/2017-present: Research Scientist @ Yahoo Research, New York, NY
06/2016-07/2016: Image Feature Team Graduate Student Lead @ Accenture, Bangalore, India
Led the Duke Student Image Feature Team on a computer vision project for Accenture's client, worked on object detection using deep learning
01/2016-05/2016: Teaching Assistant @ Duke University, Durham, NC
Teaching Assistant for graduate course Advanced Machine Learning (STA 571)
09/2015-07/2016: Accenture Fellow @ Accenture and Duke University
Our proposal "Flexible and Scalable Probabilistic Models for Healthcare Data" is one of the four selected proposals for Accenture-Duke Analytics Research Program.
08/2015-12/2015: Teaching Assistant @ Duke University, Durham, NC
Teaching Assistant for graduate course Information Theory (ECE 587/STA 563)
05/2015-08/2015: Research Intern @ Oracle Labs (previous Sun Labs), Burlington, MA
Proposed a Bayesian bilinear factor method for context-based topic learning. The method models (1) the correlation between topics, and (2) evolution of word-word co-occurrence pattern as the distance between the current word and the word in the context grows. For the inference, batch Gibbs sampling as well as online Gibbs sampling were implemented. In addition, I further developed a parallel implementation for online learning. All inference methods were implemented using Java. We applied our model on large-scale Wikipedia dataset, and reasonable results were obtained with respect to computational acceleration (due to parallel computing), word embedding, topic and topic correlation learning.
05/2014-08/2014: Research Intern @ Adobe Research, San Francisco, CA
Worked on talent search/recommendation project. The project aimed to recommend talented creatives and high-quality projects to users and company recruiters. The recommendation system was accomplished by using a hierarchical regression model, which scaled to large appreciation data containing about 2 million users, 8 million projects, and 400 million appreciation records. The model was implemented using Python, and run on Amazon EC2.