implemented the paper - time series analysis - created LSTM architecture, implemented custom Loss,
preprocessing, post-processing pipeline as well as evaluation metrics using Python and Pytorch

extract date from bills, Pre-process image (skew correction, thresholding), used tesseract
to get the text, do a regex search for date patterns

Recognize Kannada digits 0-9 - created a CNN with max-pooling, dropout, batch normalization - and
also got good results with Transfer Learning using VGG16 and RESNET (like) architecture which was pre-trained on MNIST

Contributed on GitHub - Tracking the desktop applications in real-time and time spent on
each application - Built a patch for this application to support the Linux environment