A chat-bot has been developed to provide an interactive conversation with user with capabilities of understanding, context and generation of
responses with state-of-the-art model BERT.

A model has been developed to find out the hidden topics present in the
document (GSD raised by employee) in a dataset of IT related query.
LSA, LDA, LDA2VEC approach has been used.

A model has been developed to generate a extractive summary of a given document using page-rank algorithm

A model has been developed to process forms like structure and identify
the handwritten words in it. Problem is tackled in two ways: constraint
form (separated letters) and unconstraint form (free form). Capsule net-
work and Autoencoder is used in process. Some of data is generated
using DCGAN. Model is giving 88?curacy over constraint forms. In
unconstraint, we have used LSTM–CTC architecture with the accuracy of 73.

A search engine pipeline is developing to collect data, index, rank the
page and utilizing it to provide better recommendation for a particular
search query.