Picking berries...





Contact me

Error 000

Creativity Starts Here.

Welcome To My Home Page

The site is proudly presented by Blarry

Who is this "Blarry"?

Blarry Wang

I'm a junior computer science major at Univeristy of Washington, Seattle.

My focus is on machine learning, data science and systems. But you'll find me doing QA related works, Arduino hacks, or designing and building a website mostly from scratch (this one).

In addition to my career-related interests, my other passions include political science, swing dancing, and drumming.


A Machine Learning Library

Personal Project. Built with Python and numpy. MLflow is a computation graph executor for machine learning implemented in Python. It has built in automatic differentiation and optimizers. You can implement, train, and test your neural networks in MLflow. The models can also be serialized for later use. See the project on Github.

Neural Network

Build neural network within minutes.


Automatic optimization.

ML Dependency Engine

Automatically Schedule Ops with Multithreading

School team project, built with C++, Python and tvm. The dependency engine is built to automatically schedule operations in a computation graph. The engine works with both the declarative style and imperative style of building neural networks. It provides a high level interface in Python and implements multithreading in C++. See the project on Github.


Scheduling on multiple threads.


Backend: C++ speed.

ML From Scratch

Python + Numpy = ML Algos

Personal project. Built with Python and Numpy. Implementing a neural network without the help of a library is one undertaking one should go through at least once in their lifetime… here are my attempts. See the project on Github.


Handwritten from scratch.


Include backprop and optimization.

Mark V

VR Experience Enhanced by Fans

Side team project. Built with Arduino and C#. Mark V is an Ironman-like VR flying experience, enhanced with physical wind feedbacks using four arrays of compact fans. My role focuses on the hardware hacks with Arduino: coding the Arduino to communicate with VR while controlling the fans, and setting up the external frame.


Fly in different landscapes.


Realistic, immersive experience.


Data Science Engineering Intern at TIBCO Inc.

Summer, 2018 -- Now

I worked in TIBCO Inc. Seattle office full time during the summer and is working part time during the school year.

As an engineering intern on a data science team, I get to practice CS related concepts in a data science context. My main accomplishments include building Jenkins pipelines for different platforms, creating analytics for test results, provide a terminal based solution to view analytics.

In addition, I also published two deep learning related articles during my internship. Find them here: 1, 2.

View my LinkedIn >>


Transfer Learning Using Keras and R / TERR

Aug 28, 2018

This example will the Keras R package to build an image classifier in TIBCO® Enterprise Runtime for R (TERR ™).

You will also see:

1. how to subset of the Cifar-10 dataset to compensate for computation resource constraints;

2. how to retrain a neural network with pre-trained weights;

3. how to do basic performance analysis on the models.

Read full article >>

Get in touch

Thanks for stopping by! If you want to get in touch with me, email me at blarry.personal@gmail.com.

Or, find me on Github https://github.com/BBBBlarry

LinkedIn: https://www.linkedin.com/in/blarry