Writing
Series
Topics that took more than one article, meant to be read in order. The Transformer Internals foundations are a series too.
-
Policy Optimization for LLMs: From Fundamentals to Production
From RL foundations and PPO through GRPO and GDPO to a map of the GRPO variants: policy optimization for aligning language models.
-
Adaptive Optimization at Scale: Contextual Bandits from Theory to Production
From the decision of whether to use a bandit, through regret theory and four algorithm implementations, to offline evaluation and deployment.