Using Machine Learning to build a Wordle Strategy

As an independent research project, I wanted to explore key features in ML. I made a Bayesian model, Unsupervised model and Autoencoder from scratch in Python.
My models simulate wordle attempts with different strategies and starting words to find the optimum. My best model could complete the wordle in an average of 3.6 guesses, beating the average human who does it in 4.
I even presented my findings at a research seminar (see slide 2).