A directory of my deep learning projects.

I spent my spring 2017 semester working on deep learning applications as a Solution Architect Intern at NVIDIA.

• Unsupervised Learning for Drug Discovery: Implemented Stacked Denoising Autoencoder and Deep Embedded Clustering (DEC) models in TensorFlow to help a pharmaceutical customer automate their drug discovery process.
• Domain Randomization for VR Object Localization: Experimented using 3D object localization to map real-world objects (e.g. chairs, tables) into VR. Generated synthetic sample data in Unreal Engine 4 using techniques for domain randomization, which enables theoretically unlimited training dataset size.

I research deep reinforcement learning at Brown University under Michael Littman. These are projects I've done in the lab:

• Adversarial Video Generation: A TensorFlow Implementation of "Deep Multi-Scale Video Prediction Beyond Mean Square Error" by Mathieu, Couprie & LeCun.

I mess around with neural nets in my free time to learn new things and create cool AI art:

• encore.ai: Generate new lyrics in the style of any artist using LSTMs and TensorFlow. (Winner – Best ML Hack, HackMIT 2016.)
• Autonomous Driving: I was accepted into the innaugural class of Udacity's Self-Driving Car Nanodegree Program (4.5% acceptance rate). These are projects from that course:
  • Behavioral Cloning: Used Keras and Deep CNNs to train a behavioral cloning network to predict steering angles from images based on samples of my driving.
  • Lane Detection: Used classical computer vision to detect lane lines, curvature and car position from camera images in various lighting conditions.
  • Car Detection: Used a SVM classifier on HOG, Color Histogram and Color Bin features to detect other cars on the road.