Principal Systems ML Engineer — Scale ML Infrastructure

Meta; Santa Fe (NM); Sur place; USD 219 000 - 301 000; Plein temps

Résumé du poste

Meta is seeking a principal-level Software Engineer to define architectural foundations powering large-scale ML infrastructure, spanning training systems, inference pipelines, ML compilers, high-performance computing frameworks, and on-device optimization.

You will shape multi-year roadmaps, drive performance improvements, mentor engineers, and communicate complex ML systems architecture to both technical and non-technical audiences.

Qualifications

• Bachelor's degree in Computer Science, Computer Engineering, or equivalent practical experience.
• 12+ years of experience in software engineering with deep specialization in one or more ML systems domains including AI infrastructure, ML compilers, high-performance computing, GPU architecture, ML frameworks, or on-device optimization.
• Experience architecting and delivering large-scale ML training or inference infrastructure that has had measurable impact across multiple engineering organizations.
• Experience leading multi-year cross-functional technical initiatives, including defining metrics, managing dependencies, and driving execution across organizational boundaries.
• Experience developing high-performance ML systems infrastructure in C++, Python, or CUDA, including work at the intersection of hardware and software.
• Experience influencing technical direction and engineering practices across multiple teams through written proposals, design reviews, and stakeholder alignment.

Responsabilités

• Identify and solve the most complex cross-system ML infrastructure challenges spanning training, inference, compiler optimization, and hardware-software co-design.
• Define extensible architectural standards and technical foundations for ML systems that enable consistency and reliability across multiple engineering organizations.
• Develop and own the multi-year technical roadmap for ML systems infrastructure, balancing short-term delivery with long-term platform health and competitive positioning.
• Leverage AI-native tooling and workflows as a force multiplier to eliminate entire categories of engineering toil and accelerate cross-disciplinary work across the ML systems stack.
• Drive performance improvements across large-scale ML training and inference systems by identifying bottlenecks that span multiple subsystems, ownership boundaries, and abstraction layers.
• Establish invariants, correctness proofs, and systemic reliability practices that prevent whole classes of failures across ML infrastructure pipelines.
• Partner with research, hardware, and product engineering teams to translate theoretical ML systems advances into production infrastructure that delivers measurable efficiency and capability gains.
• Assess emerging AI and computing technologies, evaluate competitive ML infrastructure trends, and influence organizational strategy to ensure technical competitiveness.
• Mentor engineers across the organization by providing customized coaching, leading engineering programs, and establishing a culture of thoroughness and high craft in ML systems development.
• Communicate complex ML systems architecture and strategy clearly to technical and non-technical audiences, producing reference-quality design documents and roadmap artifacts.

Connaissances

Cross-system ML infra

Architectural standards

Roadmapping ML systems

AI-native tooling

Performance optimization

Reliability practices

Stakeholder alignment

Formation

Bachelor's degree

Outils

MLIR

XLA

TVM

CUDA

C++

Python

Description du poste

Meta is seeking a principal-level Software Engineer to define architectural foundations powering large-scale ML infrastructure, spanning training systems, inference pipelines, ML compilers, high-performance computing frameworks, and on-device optimization.

You will shape multi-year roadmaps, drive performance improvements, mentor engineers, and communicate complex ML systems architecture to both technical and non-technical audiences.