Machine Learning Performance Modeling Architect

Meta•Sunnyvale, CA

11h

About The Position

Reality Labs focuses on delivering Meta's vision through On-device AI. The compute performance and power efficiency requirements of these workloads require custom silicon. Reality Labs Silicon team is driving the state of the art forward with breakthrough work in computer vision, machine learning, mixed reality, displays, sensors, and new ways to map the human body. Our chips will enable On-device AI assistance and features, where our real and virtual world will mix and match throughout the day. We believe the only way to achieve our goals is to look at the entire stack, from transistors to architecture, firmware, and algorithms. We are seeking talented professionals to support the development and optimization of machine learning workloads and performance modeling for custom hardware and software platforms. In this role, you will contribute to analytical and simulation-based modeling and analysis, collaborating with cross-functional teams to build scalable and efficient solutions. Ideal candidates have a strong background in machine learning, system architecture, and performance modeling, and thrive in collaborative, hands-on environments.

Requirements

Bachelor's degree in Computer Science, Computer Engineering, relevant technical field, or equivalent practical experience
8+ years of experience in IP/SoC/System performance modeling and workload analysis/optimization for low-power/high-performance accelerators
8+ years of experience with programming languages (C/C++ and Python), script automation and data visualization
Experience evaluating architectural tradeoffs in performance, power and image quality
Experience employing scientific methods to debug, diagnose and drive the resolution of complex, cross-disciplinary system issues

Nice To Haves

MS EE/CS or equivalent in relevant areas
Experience with building or modifying full-system performance simulators and analytical models
Experience with collecting and interpreting performance counters using SW profilers
Experience in machine workload development in Pytorch or similar ML toolchains
Experience with thermally constrained power/performance optimization in mobile devices
Experience operating under your own direction, driving high-level direction
Experience with telemetry generation and analysis
Experience with ML hardware architectures and use cases
Understanding of SoC and System architecture and heterogeneous compute principles
Experience Collaborating closely with the machine learning and system architecture teams to develop performance models for AR/VR machine learning software-hardware verticals
Experience supporting all phases of Silicon SoC development from a machine learning vertical modeling - from early definition on through specification, architecture and productization

Responsibilities

Lead power/performance modeling and analysis of machine learning software-hardware components and use cases
Capture machine learning workloads from applications and system usages
Support all phases of Silicon SoC development
Contribute to execution of our silicon technology / machine learning roadmap to make beyond state-of-the-art advances in performance, power consumption and form factor
Work across disciplines, build new methodologies, juggle/coordinate multiple initiatives