Staff AI Research Scientist: TPU & Large-Scale ML

Meta; Menlo Park (CA); On-site; USD 184,000 - 257,000; Full time

Benefits offered by this job

Bonus

Equity

Job summary

Meta AI Research is seeking a Staff Research Scientist to lead TPU performance optimization and large-scale model training within Meta's native PyTorch stack in Menlo Park. You will drive systems-level ML research, collaborate across research and engineering teams, and translate findings into production-ready optimizations at scale.

The role requires deep expertise in TPU kernels, memory management, and distributed training strategies, with a track record of impactful publications and mentorship.

Qualifications

• Bachelor's degree in Computer Science, Computer Engineering, or equivalent practical experience.
• 8+ years of experience in ML systems, model optimization, or HPC research.
• Experience with TPU architecture and performance optimization, including profiling, kernel development, and memory management.
• Experience with XLA compilation, graph optimization, and low-level performance tuning for accelerator hardware.
• Experience developing and optimizing large-scale distributed training systems, including parallelism strategies.
• Experience with PyTorch and its integration with accelerator backends.
• Experience communicating complex technical findings in writing, including technical reports and publications.

Responsibilities

• Lead the design and execution of TPU performance optimization research, including kernel development, memory optimization, and compute efficiency improvements.
• Develop and optimize Pallas kernels for large-scale model training and inference on TPU architectures.
• Drive model optimization techniques including MoE, tensor, and pipeline parallelism and other distributed training strategies.
• Optimize first party models within Meta's native PyTorch stack, integrating with XLA and TPU execution.
• Identify and resolve complex challenges in training efficiency, latency, and reliability with novel approaches.
• Define and drive multi-quarter research roadmaps for TPU optimization and align with organizational goals.
• Establish experimentation frameworks for benchmarking, profiling, and data-driven decisions.
• Translate research findings into production-ready optimizations with deployment and reliability at scale.
• Communicate findings and trade-offs through publications, design docs, and presentations to technical and non-technical audiences.
• Mentor researchers and engineers on TPU optimization techniques and experimental rigor.

Skills

TPU performance optimization

Large-scale model training

Systems ML research

PyTorch integration

Distributed training

Education

Bachelor's degree in CS/CE or equivalent

Tools

XLA compilation

Kernel development

Memory management

Profiling

Job description

Meta AI Research is seeking a Staff Research Scientist to lead TPU performance optimization and large-scale model training within Meta's native PyTorch stack in Menlo Park. You will drive systems-level ML research, collaborate across research and engineering teams, and translate findings into production-ready optimizations at scale.

The role requires deep expertise in TPU kernels, memory management, and distributed training strategies, with a track record of impactful publications and mentorship.