Experience
Researcher, Vemba Corp - May 2015 - Present
Vemba (Vemba Corporation, Toronto, Canda)
- Designing deep-learning methods to determine the similarity between videos, based on their semantic content
- Building convolutional neural networks (CNN) to detect salient objects in video
- Developing methods to use the hidden layer features of image-trained CNN to form an embedding method, in order to associate similar videos
- Implementing natural language processing techniques, such as TF-IDF, for mining video training data from a video library with closed captions
Researcher, Intel Intelligent System’s Group - May 2012 - Sept 2013
Intel (Intel Corporation, Toronto, Canda)
- Implemented and adapted methods for pedestrian/face recognition and tracking for use in anonymous video analytics software.
- Responsible for adapting existing literature and developed new methods in conjunction with RGBD data obtained from Microsoft Kinect as well as improving accuracy of existing age, gender and race classifiers.
- Developed Histogram of Oriented Gradients models for pedestrian recognition and tracking, as well as machine learning systems to identify demographic information, i.e. race, age, gender
- Patent: Rohan Chandra, Abhishek Ranjan, Shahzad Malik, “TECHNOLOGIES FOR INCREASING THE ACCURACY OF DEPTH CAMERA IMAGES”, International Patent number: PCT/US2013/041864
Teaching Assistant - January 2012 - May 2012
University of Toronto - CSC320: Introduction to Visual Computing
- Teaching assistant for a third year introductory visual computing course, primarily taught in C++ and MATLAB
- Responsible for conducting tutorials to supplement lectures, attending to student questions both in person and by email, as well as grading of student assignments and tests
- Topics Covered: Camera response function, image derivatives and applications, local curve analysis, edge/corner detection, Gaussian pyramids, Haar wavelets, SIFT, and Homographies
Junior Software Developer - July 2010- September 2011
CAST Software (CAST Group Inc.)