World-scale vision-language-action (VLA) models are the new frontier in AI for autonomous driving and robotics, enabling systems to perceive, reason and act in complex real-world environments. Using the open-source Pi-0.5 model as an exemplar, we will review the structure of VLA models and examine how they combine advanced techniques from both language and vision transformers into a unified architecture. We’ll then examine the unique challenges VLAs introduce for embedded deployment, including quantization complexity and a mix of compute-dominated and bandwidth-bottlenecked workloads that must be handled efficiently. We will explore practical strategies for mapping VLA workloads to embedded compute engines and present results from porting and optimizing Pi-0.5 on Quadric’s Chimera GPNPU, showing how Chimera’s combination of software control and determinism makes it uniquely well-suited to optimized VLA inference at the edge.
Porting and Optimizing Advanced Vision-Language-Action Models for Embedded Autonomous Systems
Session Speakers
-
Mike Leonard
Software Architect, Quadric
Mike Leonard is a Software Architect at Quadric, where he leads development of the graph compiler stack and programming models behind the Chimera GPNPU processor. His work bridges advanced hardware-software co-design with real-world AI deployment, enabling developers to efficiently run modern neural networks at the edge. As a lead architect of the Chimera Graph Compiler, Mike and his team have delivered automated support for hundreds of CNN and transformer models, achieving high performance through aggressive operator fusion, layout optimization and intelligent, automated memory management. Mike also created ChiPy, a Python-based, tensor-level programming model that combines the productivity AI developers expect with performance tuned for specialized AI hardware. Mike works closely with industry partners to translate emerging AI models into production-ready solutions and to advance the state of practical, high-performance edge AI. He holds BS and MS degrees in Computer Science from Northwestern University and is based in San Francisco.
