Principal M-LLM Post-Training and Execution Software Engineer, XR

Google•San Jose, CA

4d•$307,000 - $427,000

About The Position

At Google, we put our users first, and the Android XR team is building the future of how users experience the world through augmented and virtual reality. As the Principal SWE for Multimodal LLM (M-LLM) Post-Training & Execution, you will lead the critical efforts to take state-of-the-art research models and transform them into production-ready features for the XR platform. This role is uniquely positioned at the intersection of model alignment and high-performance execution. You will ensure our models are not only intelligent but also safe, reliable, and optimized for the real-time demands of XR environments. Your work will empower XR devices to "see" and "understand" the world in ways never before possible. The Platforms and Devices team encompasses Google's various computing software platforms across environments (desktop, mobile, applications), as well as our first party devices and services that combine the best of Google AI, software, and hardware. Teams across this area research, design, and develop new technologies to make our user's interaction with computing faster and more seamless, building innovative experiences for our users around the world.

Requirements

Bachelor’s degree in Computer Science, Mathematics, or equivalent practical experience.
15 years of experience in software engineering, specifically in large-scale machine learning systems or computing.
Experience driving the go-to-market strategy for AI-driven technologies.
Experience leading engineering teams through the full execution cycle of AI products.

Nice To Haves

Experience with high-performance inference engines and custom AI accelerators.
Advanced knowledge of computer vision pipelines, 3D perception, and their integration with large-scale language models.
Distinctive problem-solving and analytical skills applied to the challenges of real-time spatial computing.
Specialized expertise in large-scale post-training workflows, including Reinforcement Learning from Human Feedback (RLHF) and fine-tuning at scale.

Responsibilities

Design and scale the post-training infrastructure for Multimodal LLMs, including instruction tuning, RLHF, and automated evaluation frameworks.
Architect and optimize the execution engines that support real-time M-LLM inference on XR hardware, balancing high-fidelity output with battery-conscious processing.
Lead the development of large-scale computer vision and multimodal data pipelines to improve model understanding of spatial and environmental contexts.
Partner with Product and UX teams to define the "intelligence" primitives that will power the next generation of spatial applications.
Set the standard for model safety and alignment for XR-specific use cases, ensuring responsible AI deployment at scale.