As a full-time data scientist, there are many things that do not fall under my job description (e.g. web development and time series analysis) and many things that do (e.g. deep learning and NLP).
It's important to build new skills and to improve existing ones, whether they are relevant to your day-to-day job or not. So here are some of the coding projects I've worked on in my spare-time.
"Try to learn something about everything and everything about something." - Thomas H. HuxleyAn intuitive guide to understanding RLHF in the context of training LLMs.
Neural style transfer allows you to merge the style and contents of 2 different images together. See how I made turned my portrait into a Picasso painting.
Do you need a new book to read? Why not try this book recommender app I built using Flask? It's really easy, just enter a few books you really like and wait for your results!
Knowledge Distillation involves a larger pre-trained model (teacher) and a small (student) model that learns to mimic the behaviour of the teacher model. See how this can be used to reduce to size yet maintain the performance of language models.