I built a Python library that parses MIDI files in bytes and plays them as audio. For this project, I employed the MIDI file specification to decode the file into audio data. To play the audio, I used signal processing techniques and frequency modulation to generate the audio signal that was sent to the device. Finally, I used the Python packaging system to deploy TiMIDIty to the Python Package Index.
I trained a convolutional neural network on the MNIST digit dataset in TensorFlow and saved the model for web deployment. Using the Tensorflow.js and React, I created a small web application that recognizes user-drawn digits on an HTML canvas.
Vim Docs is a realtime collaborative web editor that uses basic keybindings from the vim text editor and collaboration like Google Docs. I built the site using a Vue frontend and an Express.js backend built with Node.js. I used Heroku to deploy the website and maintain it.
TFCP is a code generation AI that harnesses the large code repositories of competitive programmers and trains a recurrent neural network on their code. Using Tensorflow and Keras, I created a Python script that would learn someone's coding style and generate similar pieces of "code" that mimics their way of programming.