Staff Applied Machine Learning Software Engineer - Generative AI

AMD - San Jose, CA

Full-time - Mid Level

San Jose, CA

Computer and Electronic Product Manufacturing

About the position

At AMD, we are committed to transforming lives through our technology, enriching industries, communities, and the world. Our mission is to create exceptional products that accelerate next-generation computing experiences, serving as the foundation for data centers, artificial intelligence, PCs, gaming, and embedded systems. The culture at AMD is built on pushing the boundaries of innovation to tackle the world's most pressing challenges. We prioritize execution excellence while fostering a direct, humble, collaborative, and inclusive environment that values diverse perspectives. The AI Group at AMD is on the lookout for talented and motivated engineers and scientists to contribute to Generative AI inference solutions. In this role, you will delve into and enhance state-of-the-art research from both academia and industry, innovating in software development, model optimization, and compression algorithms for Generative AI applications, including large language models (LLMs), Stable Diffusion, and multi-modal models. As a member of the core AI Solutions team, you will enable Generative AI across a wide array of AMD products, including RyzenAI, XDNA/NPU, CPU, GPU, and embedded architectures. Your contributions will significantly shape and influence AMD's inference and training solutions across the software stack, from frameworks like PyTorch and ONNX to custom hardware such as NPU and GPU. If you are passionate about software engineering, keeping up with the latest literature, generating novel ideas, and leading their implementation through high-quality code, this role is ideal for you. You should possess a deep understanding of foundational AI model architectures, software development, and AI/ML frameworks such as PyTorch, along with expertise in quantization and sparsity algorithms. Demonstrated experience in efficiently mapping AI models onto GPUs and/or other hardware accelerators is essential. You will thrive in a dynamic, fast-paced development environment, showcasing excellent collaboration skills and a spirit of innovation while working with multiple geographically dispersed engineering teams. Your work will involve developing next-generation framework software and guiding other senior developers and domain experts.

Responsibilities

• Accelerate inference of Generative AI on AMD's products.
• Develop tools and techniques for model analysis, profiling, performance projections, and analyzing architecture bottlenecks.
• Architect and prototype custom kernels on GPUs and CPUs (HIP, CUDA, OpenCL, Triton, etc.).
• Optimize deep learning inference pipeline including graph compilation using AMD AI compilers.
• Reproduce and improve upon state-of-the-art quantization, pruning, optimization algorithms in PyTorch and Python.
• Develop high-quality software to enable next-gen solutions.
• Collaborate with AI/ML frameworks and infrastructure teams to enable new algorithms in the platforms.
• Collaborate with AMD Research and Architecture teams to improve future products.
• Act as a team player and work with a geographically distributed team.
• Influence the direction of AI/ML platforms for inference and training.

Requirements

• Deep understanding of foundational AI model architectures.
• Proficiency in software development and AI/ML frameworks such as PyTorch.
• Experience with quantization and sparsity algorithms.
• Demonstrated ability to efficiently map AI models onto GPUs and/or other hardware accelerators.
• Strong development and debugging skills in Python.
• Experience in C++ programming for GPUs and/or custom accelerators.
• Solid understanding of CNN, Transformer model architectures, LLMs, and Stable Diffusion models.
• Knowledge of LLMs/LMMs fine-tuning methods like RLHF.
• Knowledge of parameter-efficient techniques like LoRA and QLoRA.
• Contributions to open ML research or developer community.
• Strong communication and problem-solving skills.
• Academic credentials: BS, MS, or PhD in Computer Science, Computer Engineering, Electrical Engineering, or related technical fields.

Nice-to-haves

• Familiarity with torch.fx and/or other AI compilers, execution runtime.
• Highly preferred - knowledge of custom accelerator hardware.

Benefits

• Base pay depending on skills, qualifications, experience, and location.
• Eligibility for incentives such as annual bonuses or sales incentives.
• Opportunity to own shares of AMD stock and discounts when purchasing AMD stock through the Employee Stock Purchase Plan.
• Competitive benefits package.