Product

 

Smart Mirror: Snow White

Course project for ME 216M (Intro to Design of Smart Product) at Stanford. A mirror that displays jokes to you until you smile. Won “most innovative” award in class.

RelateIQ Gmail Integration

Full time project at RelateIQ – a Chrome extension that augments CSM/recruiter/relationship managers’ gmail workflow. It provides augmented contact profile, intelligent follow-ups suggestions, close-loop management, and read receipts without needing to leave your inbox.

ALittle

Awesome bookmarking/todo-list tool that tracks progress of webpages/videos

Research

 

3D Reconstruction with End-to-End Viewpoint NN

Course project in CS 231A (Computer Vision) at Stanford – an end-to-end neural network for 3D reconstruction. The network takes 2D images as input, and produces a 3D voxel grid (think pixels in 3D) as output. State of the art result. (68.1% Intersection-over-Union)

DeepShuai: Deep Reinforcement Learning based Chinese Chess Player

Course project in CS 234 (Reinforcement Learning) at Stanford – a Chinese Chess agent trained via Reinforcement Learning.

Multitask Learning and Extensions of Dynamic Coattention Network

Course project in CS 224N (Natural Language Processing w/ Deep Learning) at Stanford – given a paragraph from wikipedia and a question about it, use deep learning to produce an answer.

Iterated Prisoner’s Dilemma with Reinforcement Learning

Course project in PSYCH 209 (Neural Network Model of Cognition) at Stanford – train an agent to play Iterated Prisoner’s Dilemma using Reinforcement Learning.

Predicting Hit Songs with MIDI Musical Features

Course project in CS 229 (Machine Learning) course at Stanford – This paper predicts hit songs based on musical features from MIDI files. 88% precision (30% recall) achieved.