Member of Technical Staff, Frontiers of Deep Learning Scaling

Own end-to-end experiments to scale intelligence by scaling compute effectively in large-model pretraining.
Define what to scale up—data quality, continual learning, unified modalities, or novel model architectures—and pursue viable scaling paradigms.
Design and execute large-scale, long-duration training campaigns with stability and efficiency at the core.
Collaborate across data preparation, evaluation, and experiment iteration to drive rapid, reliable progress.
Develop distributed training strategies on multi-GPU hardware using Python, JAX, PyTorch, and Rust.
Communicate findings concisely, maintain rigorous experimentation discipline, and own outcomes from idea to impact.