Software Engineering Manager, AI Kernels & Performance Optimization — MTIA Software

Meta•Menlo Park, CA

2d•$184,000 - $257,000

About The Position

Meta is seeking a Software Engineering Manager to lead the AI Kernels and Performance Optimization team (KORE) within the MTIA Software team. MTIA is Meta’s family of in-house AI accelerator ASICs, used for production recommendation and ranking workloads and expanding into generative AI inference and training. The MTIA Software team builds the accelerator software stack, including its compiler and LLVM toolchain, runtime, kernel authoring frameworks and libraries, developer tooling, and deep PyTorch integration, in collaboration with silicon teams across successive generations. KORE develops high-performance compute and communication kernels, including GEMM, attention, normalization, collectives, elementwise and reduction fusions, and sparse and quantized paths. The team builds kernel libraries with broad PyTorch operator coverage, C++ and Python kernel authoring frameworks and DSL surfaces, and production kernels for new architectures before first silicon. Its work targets roofline-level performance and programmability at scale across features such as matrix engines, on-chip reduction fabrics, and software-managed memory hierarchies. The manager will own workload performance, drive kernel development and optimization strategy, and build a team delivering high performance on MTIA accelerators. The team profiles and diagnoses performance across the stack—from instruction scheduling, memory hierarchy and DMA behavior, and on-chip interconnect to multi-device collectives—and drives fixes at the appropriate layer. The role partners with silicon architecture teams on hardware/software co-design, measures proposed feature value with real kernels, and advocates for software-stack needs. It is a hands-on technical leadership role involving direct contributions to architectural decisions as well as enabling the team to deliver effectively.

Requirements

At least 4 years of software engineering experience focused on systems-level software, platform infrastructure, or distributed systems.
At least 2 years of experience managing or leading engineering teams, including setting technical direction and supporting engineer growth.
Experience driving technical roadmaps and executing against them with measurable outcomes in ambiguous or high-complexity environments.
Experience identifying and resolving engineering efficiency issues such as on-call burden, technical debt, or gaps in automation.
Experience collaborating across product, design, and data functions to deliver software with business and user impact.

Nice To Haves

Track record of hands-on technical contribution as a manager, including authoring design documents, reviewing system architectures, or writing production code.
Experience with hardware/software co-design cycles, including collaboration with silicon architects, validation of architectural features with real kernels, and influencing hardware roadmaps using software performance analysis.
Experience with customized AI accelerators such as TPUs, AI ASICs, or similar non-GPU architectures, with direct involvement in kernel development and roofline enablement.
Familiarity with transformer and attention kernel design, including FlashAttention-class algorithms, KV-cache management, paged and chunked attention, MoE routing, and expert dispatch.
Experience with distributed execution and collective communication, including NCCL/RCCL-class primitives, tensor and expert parallelism, and overlapping communication with compute.
Demonstrated ongoing AI skill development, such as prompt/context engineering or agent orchestration, and keeping current with emerging AI technologies.
Experience managing or leading teams working on high-performance kernel development, accelerator software, HPC, or ML systems performance.
Hands-on kernel development for novel hardware, including pre-silicon bring-up on simulators or emulators and first-light optimization on new silicon.
Experience building or contributing to high-performance kernel libraries or frameworks such as CUTLASS, cuBLAS, cuDNN, Triton, Helion, oneDNN, Composable Kernel, or comparable internal equivalents.
Demonstrated ability to integrate AI tools to optimize or redesign workflows and produce measurable impact, such as efficiency gains or quality improvements.
Experience with compiler and code-generation technologies relevant to kernels, such as MLIR, LLVM, TVM, XLA, Halide, or polyhedral scheduling.
Experience with low-precision numerics and quantization, including FP8/E4M3/E5M2, MX and other block-scaled formats, and INT8/INT4, including error analysis and calibration.
Experience adhering to and implementing responsible, ethical AI practices, such as risk assessment, bias mitigation, and quality and accuracy reviews.

Responsibilities

Lead and manage a team of systems engineers delivering multiple concurrent projects across ambiguous, high-impact areas of Meta’s platform infrastructure.
Create roadmaps and technical strategy for systems-level software, aligned with broader product and engineering goals.
Contribute directly through architecture reviews, design documents, and hands-on coding to maintain technical credibility and unblock the team.
Identify and resolve systemic issues affecting team velocity, including on-call load, technical debt, and gaps in tooling or automation.
Partner with product, design, and data science to support cross-disciplinary engineering and outcomes that advance business metrics.
Set goals and expectations for engineers, provide timely and actionable feedback, and proactively address performance gaps.
Manage staged rollouts and feature delivery using feature flagging and A/B testing frameworks while ensuring engineering quality and system reliability.
Recruit, onboard, and develop engineers, building a team structure that minimizes single points of failure and scales with organizational needs.
Communicate team goals, technical decisions, and progress to cross-functional partners and leadership, including through written design proposals and postmortems.
Contribute to organization-wide engineering efforts by sharing workflows, tooling practices, and systems engineering patterns.
Own the performance of workloads running on MTIA accelerators and drive kernel development and optimization strategy.
Guide profiling and performance root-cause analysis across instruction scheduling, memory hierarchy and DMA behavior, on-chip interconnect, and multi-device collectives; drive fixes to the appropriate software or hardware layer.
Partner with silicon architecture teams on hardware/software co-design, quantify proposed feature value with real kernels, and advocate for software-stack changes.