At MIT Laboratory for Information and Decision System - Prof. Cathy Wu LabImitation learning and Generative Model for modeling vehicle-to-vehicle interaction, trajectory prediction, and traffic simulator.

@ Media Labs, Camera Culture Group. (Prof. Ramesh Raskar)Working on computer vision, computer graphics, Non-of-sight-imaging (NLOS). Multi-bounce LiDAR returns to map the environment, especially handling specular/reflective objects.

On the AR/VR team, research on Place Recognition problem and machine learning methods for relocalization.

Intention prediction (car cut-ins, cut-outs), trajectory prediction.

(Company was later acquired by Magna International)- Re-architecture perception and Machine Learning stack- 3D object detection, tracking

- At level 5 - work on self-driving car, mainly on Prediction problems.- Deep Learning modeling, system integration.

Uber ATC (Advanced Technologies Group)- Deep Learning, self-driving car

- Autonomous self-driving car - Robotics, Computer Vision, 3D lidar object detection- Machine Learning, deep learning, - Convolutional Neural Network for dense scene segmentation and pixel-wise classification

(Prof. Chieh-Chih (Bob) Wang)- Learned about SLAMMOT (Simultaneously Localization and Mapping + Moving Object Tracking) for robotic autonomous navigation system in dynamic environment. - Used deep learning, namely Reconstruction Independent Component Analysis (RICA) and Recurrent/Recursive Neural Network, to learn spatio-temporal features to enhance moving objects detection in stereo video image, such as pedestrians, cars, bicycles in traffic scenes.

(Dr. Thomas Lu)- Helped develop a lightweight and fast ATR system for autonomous vehicles to detect and locate various targets such as landing sites, spacecrafts, jets, boats, in real-time video. - Worked with NASA and DoD, scan an image with optical grayscale correlator to identify region-of-interests (ROIs). - Extracted features from ROIs, using various techniques such as Principal Component Analysis, Haar-like, and texton filters - Developed classifiers to classify the features of each ROI. Design radial basis neural network for small object detection. Also developed adaboost to boost an ensemble of decision trees and neural networks to classify underwater man-made object. Both methods have improved accuracy over single neural networks previously used by the original ATR.

(Prof. Serge Belongie)-Worked with the Scripps Institute of Oceanography to make computer vision recognition tools for large scale monitoring of coral reefs.- Designed segmentation software to quickly label and identify bleaching area of coral reefs for health assessment. Developed under Matlab and C, user can segment multiple layers with foreground and background labeling, using algorithms growCut, random walker, and graphCut.