Principal Engineer, Efficient GenAI

What you'll do

• Propose and apply innovative techniques for transformer architectures, parallelism strategies, and inference optimization like speculative decoding.
• Implement novel, efficient architectures for Generative AI models to showcase performance benefits on AMD platforms.
• Integrate AMD-optimized models and libraries into open-source frameworks such as PyTorch, JAX, vLLM, and SGLang.
• Publish training recipes and promote technical work at major industry conferences and external venues.
• Co-optimize end-to-end performance for Generative AI on current and future AMD hardware and software solutions.
• Increase the adoption of agentic workflows to optimize and deploy Generative AI applications at scale.
• Collaborate with internal researchers and academic partners to drive innovation on AMD platforms.

What we're looking for

• PhD or master's degree in computer science, Electrical Engineering, Mathematics, or a related field.
• Deep technical understanding and hands-on experience with LLMs, 3D World and Action Models, or image/video generation models.
• Experience training models at scale and developing efficient approaches for distributed training and inference on AMD devices.
• Several years of experience in AI, deep learning, and related software development (preferred).
• Strong technical expertise in Generative AI model training and inference with familiarity in PyTorch, JAX, vLLM, SGLang, and MuJoCo (preferred).
• Expertise and publications in efficient model architectures, optimized training, parallelism strategies, or low-precision training (preferred).
• Experience productizing Generative AI models and training foundation models at scale (preferred).
• Excellent written, verbal, and presentation skills to coordinate internally and externally (preferred).