Efficient Model Design
Nov 6, 2026
·
1 min read
My work in neural architecture search focuses on training-free search, hardware-aware model design, and efficient architectures for vision and language models.
Tiny-NAS is a distributed, full-stack framework for zero-shot neural architecture search, with hardware-latency prediction for GPU and mobile deployment. Zen-NAS is a fast, training-free method for finding models that balance predictive accuracy and inference efficiency.
Authors
Senior Principal Applied Scientist
I lead applied research and engineering at the intersection of robotics,
reinforcement learning, agentic AI, and efficient AI systems. My work spans
embodied intelligence, large-scale model training and reasoning, inference
optimization, and deploying efficient models from cloud infrastructure to
edge devices.