Performance Modeling Engineer
Meta

Job Info

We are currently seeking a machine learning performance modeling engineer to support the development of a custom machine learning software/hardware verticals and use cases optimized for AR/VR systems. As part of the machine learning performance modeling team, you will help build performance modeling and analysis toolchains for machine learning workloads. You will work in close collaboration with the groups developing the hardware, system software and SoCs for AR/VR devices.

Performance Modeling Engineer Responsibilities:

• Lead power and performance modeling of IP components and use cases for SoC chips.
• Extract system-usage behavior of workloads and design micro-benchmarks.
• Lead thermal measurement, performance bottleneck analysis and power characterization.
• Perform modeling at the right level of abstraction given model purpose.
• Partner across disciplines to problem solve, build new methodologies and coordinate multiple initiatives.
• Ability to operate in a cross-functional environment.

Minimum Qualifications:

• Experience with programming (C++, SystemC-TLM), scripting (Python).
• Experience with power concepts, trade-offs and low power design principles.
• Experience with performance profilers and power measurement tools.
• Familiarity with Instruction Set Simulators (ISS), optimized assembly-level kernels (e.g., ARM, Tensilica).
• Experience with SoC Architecture, NoCs, memory subsystems, and heterogeneous compute principles.
• Experience creating and optimizing machine learning workloads.

Preferred Qualifications:

• Bachelors degree in EE, CS or equivalent experience.
• 3+ years of experience with C++ programming.
• 2+ years of experience with SystemC/TLM2 programming.
• Experience with CV/ML algorithms.
• Computer architecture experience, including CPU and domain-specific/ML accelerators.
• Familiarity with power management principles, thermal and di/dt analysis, multi-voltage designs, DVFS and UPF concepts.
• Experience with thermally constrained power/performance optimization on embedded devices.
• Experience with bare-metal programming, micro-benchmarking, etc.
• Experience with gathering and interpreting performance counters using tools like perf, VTune, etc.
• Experience in building performance models for custom accelerator and SOC pipelines.

About Meta: