The bot is deployed on a raspberry pi that lives out in the chicken coop. It's an image classification model built on tensorflow 2.3.1 from a library of about 10,000 images that were scraped and cleansed (by hand) from public sources.

It runs on a series of docker containers deployed on the above pi including tensorflow (to train the model periodically), tensorflow/serving, python flask (for an API), and PRAW to connect to reddit to listen to mentions of its own name. In addition to the deep learning being done with Tensorflow, bawkbawkbot also is learning through immersion by living in the coop amongst the chickens.