I'm Kushal. My research focuses on improving the systematic generalization, oversight and understanding of the training processes of large language models.

Previously, I worked on the science of neural network training and on improving the Transformer architecture at Stanford NLP, and worked on RL at Applied Compute and Prime Intellect.

Reach out if you'd like to chat about anything!