AI for Impact

Code

 

Featured Github Projects

Below are my implementations of various Deep Learning papers (Github repo for other interesting implementations). I also have useful libraries for Swift, Python, Objective-C and rapid product development.

 
 
Screen Shot 2018-01-10 at 11.53.45 AM.png

Test tube

Test tube is a python library to track, optimize and parallelize Deep Learning experiments across GPUs and compute clusters. It's framework agnostic and is built on top of the python Argparse API for ease of use.


 Image source:  keon.io

Image source: keon.io

deepRLHACKS - My notes on john schulman's talk from uc berkely deep rl bootcamp (aug 2017)

Hacks for training RL systems from John Schulman's lecture at Deep RL Bootcamp (Aug 2017)


val_epoch_0_batch_14.png

 Image-to-Image Translation with Conditional Adversarial Networks (pix2pix)

Original paper

This particular GAN (General Adversarial Network), learns a mapping between an image X to an image Y, where X is a different representation of Y. In their paper, Isola et al use this GAN to map from building facades to building images.

 

Screenshot 2018-10-03 08.20.45.png

Efficient Natural Language Response Suggestion for Smart Reply

Original paper

Tensorflow implementation of a siamese feedforward architecture to score potential responses to a conversational context.