ISRO's Astrosat records solar activity in lightcurve files. That data was cleaned and used along with consensus among multiple algorithms to detect and confirm solar flares.
This notebook contains a GAN implementation for MNIST, CIFAR and other datasets. It also contains a custom NN implementation on Keras. Click to view the notebook.
I attempted to use news sentiment about a company to predict and trade its shares in the stock market. I used keyword based news extraction form news sites, and tried using embedding matrices to perform sentiment recognition.
Used bi-directional LSTM and a convnet to try to suppress the voices in songs, so only the music could be audible. We use pydub and scipy's wavfile to edit the audio and Short Term Fourier Transform to extract features and convert it into training data.
To make a matching algorithm to pair up accountability partners. It ingests data points and outputs a list of matches (users).
In this one about fetal health, I classify the data using Decision Trees, Random Forests and Support Vector Machines. Click to view the notebook.
I did most of my data science work during interns and work. But if you look for it, you'll find a lot in my GitHub too. Have a look!