About the job
We are seeking an exceptional Senior Fellow in Engineering to serve as a technical leader for Machine Learning workload performance and optimization across frameworks and model architectures. This is a highly visible individual contributor (IC) role with broad technical scope and company-wide impact, operating at a level equivalent to Senior Director while focusing primarily on deep technical leadership rather than organizational management.
Responsibilities
Define and drive the technical strategy and roadmap for ML model optimization across AMD platforms.
Serve as the highest-level technical authority in performance optimization, guiding architecture, and implementation decisions.
Lead performance tuning, profiling, and analysis of large-scale models (LLMs, diffusion, multimodal, RecSys, generative AI) across single-node and distributed environments.
Drive hardware-software co-design initiatives to influence future GPU architectures and system-level optimizations.
Collaborate across engineering, research, and customer-facing teams to deliver best-in-class performance.
Develop advanced methodologies, tools, and infrastructure for performance estimation, modeling, and benchmarking.
Provide technical mentorship to senior engineers and influence best practices across the organization.
Represent AMD in external technical forums, benchmarks, and customer engagements.
Communicate complex technical findings and recommendations to senior leadership and stakeholders.
Qualifications
Minimum
No minimum qualifications listed.
Preferred
Sr. Fellow level of experience in performance engineering, ML systems, or related domains, with demonstrated technical leadership at scale.
Deep expertise in performance analysis, modeling, and hardware/software co-optimization.
Proven impact on optimizing large-scale ML workloads on modern accelerator architectures.
Strong ability to influence cross-functional teams without direct authority.
Experience contributing to industry benchmarks, open-source ecosystems, or published research is a plus.