David McAllister

I've studied scalable image synthesis with flow models, developing methods to train them across decentralized compute clusters¹ and to improve their RL post-training².

I've since applied these techniques to robot learning, including reinforcement learning with flow matching³ and learning context-aware humanoid locomotion from monocular video⁴ (CoRL Best Student Paper). Now, I'm training vision-language-action models on large-scale demonstration data.