A medium article I wrote on building multimodal llms from scratch in colab

An O(log n) optimization I found to speed up tokenizer throughput

Boosting OCR performances of open-source vision language models with synthetic datasets

Playing around with bayesian priors and deep learning