Ryan Liu

Software Engineer at Veeva Systems

I am currently a Software Engineer at Veeva. I graduated from UC Berkeley with a B.S. in Electrical Engineering and Computer Science (EECS). Some of my interests include robotics, machine learning, badminton and table tennis.

Projects

Deep Reinforcement Learning Model for Lunar Landing
  • Developed a reinforcement learning agent to solve the lunar lander OpenAI environment using Deep Q Learning with memory replay to successfully land the lander in at a specific location
Scalable Credit Card Fraud Detection with Kubernetes
  • Developed a DNN model to perform fraud detection for credit card transactions, using SMOTE to handle the imbalanced dataset with an AUC of 0.971
  • Developed a RESTful API with Flask and deployed with Kubernetes and Docker to support performant and scalable model serving
Dog Breed Identification
  • Developed a CNN model in PyTorch to determine dog breed
  • Deployed a real time inference pipeline and website for predictions using AWS Sagemaker and Flask
  • Also deployed to Heroku and can be tried out below. Note that it may take several seconds for a response.
Exercise Pose Detector
  • Developed a CNN model in PyTorch to determine exercise based on pose with 93% accuracy
  • Deployed a real time inference pipeline and website for predictions using AWS Sagemaker and Flask
Weather Wear
  • Developed an Android app utilizing GCP Vision API to detect articles of clothing
  • Implemented an algorithm combining weather data, past outfits and user feedback to provide personalized clothing suggestions
  • See our Calhacks post for additional information.
Feature Extension for Pintos OS
  • Extended an OS in C, implementing thread scheduling, running user programs, a filesystem buffer cache and extensible file system with a team of 4
RISC V ISA Emulator
  • Designed a RISC V Emulator capable of executing assembly code and reporting machine code and contents of registers.
  • Allows for debugging of assembly programs
Voice Controlled Robot Car
  • Built and designed a microphone signal processing circuit and developed firmware in C for the closed loop control of movement and voice recognition
  • Built a voice command controlled car using PCA and k-means for voice recognition
Housing Price Analysis
  • Visualized housing data for feature engineering using seaborn and sklearn
  • Designed a linear model with lasso regularization to predict housing prices, scoring in the top 10% of the class
Wizard Solver
  • Modeled and solved an open-ended, NP-hard constraint satisfaction problem using simulated annealing and local search in Java with a 3-person team
  • Tied for 1st place out of 300+ entries in final competition
Secure Cloud Store
  • Built a high-performance, cryptographically secure cloud storage system in Go with sharing and access controls
Yelp Maps
  • Predicts a user's rating for restaurants using previous resturant ratings and reviews
  • Presents a visualization of restaurant ratings using a Voronoi diagram where each section is colored based on the predicted rating of the restaurant in the section
Pokemon Mystery Dungeon Clone
  • Designed and implemented a clone of Pokemon Mystery Dungeon in Java
  • Involved procedural dungeon generation to randomly generate the dungeon, enemies and various items throughout each dungeon.
  • Built enemy bots capable of moving around level and attacking the player
  • Github Link