Eric Hou      (925) 353-7416      LinkedIn      GitHub      Email

University of California, Berkeley
B.A. Computer Science
Class of 2020 Regents' and Chancellor's Scholar (Top 1.5% of Class)
Research Assistant at Renewable and Appropriate Energy Laboratory


Projects

Twitter Post Generator (In Progress)
Torch, Lua, Python, Django, HTML, CSS, Bootstrap, PostgreSQL

  • App that learns to generate Tweets in users' styles with a machine learning algorithm
  • Assembling LSTM architectures to featurize character sequences in the same space as standard vocabulary to parse acronyms and internet slang
  • Building generative neural network inspired by Seq2Seq and Label Extraction and encoder-decoder LSTM models
  • Designing user-friendly interface and webapp to allow bot creation in any user's style
  • Featurizing user style and applying them to the state vector of the Twittersphere to generate new twitter posts
  • Repo: work in progress!

Scraper - Python Web Scraper
Python, Requests, Multithreading

  • A multi-threaded web scraper written in Python that allows for large-scale distributed scraping
  • Allows for lightweight user-defined scraping patterns and recursive, complex scraping patterns
  • Repo: Scraper

Synapse - CNN Generator API and Educational Tool (In Progress)
Torch, Lua, Python, Java, Processing, ControlP5

  • 2nd place hack at CSUA hackathon
  • App intended to educate people without ML experience about Convolutional Neural Networks
  • Allows users to explore the effect of hyperparameters on CNNs and see relevant data about different stages of training saved over time.
  • API allows dynamic generation of black-box CNNs with built-in training, validation, and testing scripts.
  • Implements many preprocessing steps such as normalization, edge detection, size standardization, and more.
  • Repo: CNN

hmwk.io
Java, MongoDB, Angular, Express, Node, Maven, AWS, OpenCV, Tesseract OCR

  • Hack submission for CalHacks 3.0, intended to allow users to search for factual information in an image
  • Implemented image flattening, normalization, gaussian blur, edge detection, and other image processing algorithms in Java
  • Designed and implemented text isolation algorithm to reduce noise and improve OCR
  • Integrated application with OpenCV, Tesseract, and Watson API for OCR and NLP search tasks, image entity recognition, and concept extraction
  • Utilized probabilistic model with similarity detection to improve document search built on top of the Bing and Watson APIs
  • Repo: hmwk.io

Object-Oriented Neural Network (no APIs)
Java, Java Language Features

  • A modular neural network that can be configured to tackle different problems with a container class Brain for simple customization.
  • Machine learning framework intended to provide Java access to simple neural networks
  • Implemented modular neural network framework using only Java language features
  • Utilized Java 8’s lambdas, ArrayLists, and Object-Oriented design to implement feed-forward, loss calculation, and the back-propagation algorithm
  • Repo: AI

JSON Server-Client REST API (Wrote this before I found out about Node.js)
Java, Servlets, Apache Tomcat, Google GSON, Maven

  • Event-driven API for web communication written in Java with Servlet API and Apache Tomcat with Maven for build automation
  • Utilized event loop with asynchronous events and callbacks to serve content dynamically in webapps with embedded Tomcat
  • Implemented client push and UDP-based LAN discovery for easier client-server interaction
  • Worked with locks and control flow with multi-threading to ensure scalability
  • Repo: Framework