Waymo World Model, Foundational World Model, Simulation Realism

Building AI to understand human behavior so we can drive with Waymo safely.

• Focus on Foundational Behavior Models, Foundation World Model

• Synthetic Data Generation, Distillation, Inference Techniques for Large Foundational Models

Applied Researcher with >5.1k citations in AI.

FAIAR: Engineering lead in the PyTorch Ecosystem at Meta. This includes training (PyTorch Lightning) and domain undestanding frameworks and libraries (multimodal, audio, text, vision, etc.).

Applied Researcher with >2.3k citations and experience across production training systems.

Visiting faculty at North Carolina A&T University, the largest public historically black university in the US with the largest number of black-engineering graduates.

I led the course design at the undergraduate/graduate level for advanced algorithms, machine learning, and artifical intelligence units as part of Meta's Engineering in Residence Program.

Focused on research at the intersection of content understanding for cybersecurity at the AI/Cybersecurity lab.

Working on building an industry leading multimodal content understanding platform across Facebook AI, consolidating content understanding and ranking onto a single training stack based on PyTorch to increase research to production velocity. Migrated 100+ models from Caffe2 to PyTorch.

Leading efforts to generalize advanced model compression techniques (such as quantized-aware training, post-training quantization, model distillation) across domains and tech stacks to decrease model inference latency in production.

Deployed libraries across open-source and internal frameworks such as PyTorch Lightning to improve ML data read performance by 19% (as measured by data load times).