GPU Energy Modeling Engineer for ML-Driven Architectures

NVIDIA; Santa Clara (CA); On-site; USD 116,000 - 219,000; Full time

Benefits offered by this job

Equity

Benefits

Job summary

NVIDIA is seeking an Architecture Energy Modeling Engineer in Santa Clara to join the Power Modeling, Methodology and Analysis Team. You will help build ML-based power models to reduce GPU energy consumption and collaborate with architects, ASIC Design, performance and hardware teams to study energy efficiency across NVIDIA GPUs and Tegra SOCs.

You will train and refine models, integrate energy models into architectural simulators and RTL platforms, and work to bridge early energy estimates to

Qualifications

• Pursuing or recently completed a MS or PhD in Electrical Engineering, Computer Engineering, Computer Science or equivalent experience.
• Strong coding skills, preferably in Python, C++.
• Background in machine learning, AI, and/or statistical modeling.
• Background in computer architecture and interest in energy-efficient GPU designs.
• Familiarity with Verilog and ASIC design principles is a plus.
• Ability to formulate and analyze algorithms, and comment on their runtime and memory complexities.
• Basic understanding of fundamental concepts of energy consumption, estimation, and low power design.
• Desire to bring quantitative decision-making and analytics to improve the energy efficiency of our products.
• Good verbal/written communication and interpersonal skills.

Responsibilities

• Work with architects, designers, and performance engineers to develop an energy-efficient GPU.
• Identify key design features and workloads for building Machine Learning based unit power/energy models.
• Develop and own methodologies and workflows to train models using ML and/or statistical techniques.
• Improve the accuracy of trained models by using different model representations, objective functions, and learning algorithms.
• Develop methodologies to estimate data movement power/energy accurately.
• Correlate the predicted energy from models built at different stages of the design cycle, with the goal of bridging early estimates to silicon.
• Work with performance infrastructure teams to integrate power/energy models into their platforms to enable combined reporting of performance and power for various workloads.
• Develop tools to debug energy inefficiencies observed in various workloads run on silicon, RTL, and architectural simulators. Identify and suggest solutions to fix the energy inefficiencies.
• Prototype new architectural features, build an energy model for those new features, and analyze the system impact.
• Identify, suggest, and/or participate in studies for improving GPU perf/watt.

Skills

Python

C++

Machine Learning

Verilog

Education

MS/PhD in Electrical Engineering, Computer Engineering, Computer Science

Job description

NVIDIA is seeking an Architecture Energy Modeling Engineer in Santa Clara to join the Power Modeling, Methodology and Analysis Team. You will help build ML-based power models to reduce GPU energy consumption and collaborate with architects, ASIC Design, performance and hardware teams to study energy efficiency across NVIDIA GPUs and Tegra SOCs.

You will train and refine models, integrate energy models into architectural simulators and RTL platforms, and work to bridge early energy estimates to