I am interested in how learning algorithms change model behavior, especially in large language models and reinforcement learning systems.
Right now, I am studying LLM learning dynamics, building RL agents for games, and improving the engineering habits needed to run clear, reproducible ML experiments.
- Learning dynamics of large language models, especially how SFT, DPO, RL, and related methods shape model probability distributions
- Reinforcement learning, with an interest in applying RL systems to games
- Training and inference optimization for deep learning models
- Exploring possible research topics in LLM learning dynamics
- Building a Pokemon TCG bot with reinforcement learning
- Re-learning reinforcement learning from first principles using JAX
- LinkedIn: belati-jagad
- X: @belatijagad