MODELER FOR COMPUTE OPTIMIZATION Job in Santa Clara, California US
MODELER FOR COMPUTE OPTIMIZATION #1472656
RESPONSIBILITIES:
- The primary responsibility of this job is to quantitatively model, analyze, and optimize the performance of NVIDIA's compute farm. This will include:
- Developing an adequate understanding of the engineering work flows that invoke the farm compute services in the form of jobs;
- Characterizing, classifying and modeling the nature of different types of jobs in terms of their demands on all farm resources, including compute cores, storage, network, and software licenses;
- Developing models of job demand distribution by job type through all stages of the product development cycle ? these models should allow changes in job demand distribution and resource allocation to be tested to validate throughput and efficiency of farm resources;
- Characterizing, classifying, and modeling the dynamic nature of compute cores, storage, network, and other resources;
- Analyzing the performance characteristics of alternate policies for resource allocation and scheduling in the farm;
- In conjunction with the Infrastructure team analyzing the choice of different compute and storage architectures and products;
- Developing recommendations for improving the throughput, wait time, utilization, and total cost per job of the farm;
- Helping drive the recommendations to implementation through the operations teams.
Many of the above elements would require the close collaboration and team work with other modelers and data analysts within the department.
MINIMUM REQUIREMENTS:
- Graduate level or equivalent degree in Operations Research, Operations Management, or Mathematics with specialization in quantitative (analytical and simulation) modeling, analysis, and optimization of complex, discrete, dynamic systems
- Experience in practically applying the skills to large scale discrete systems; 3+ years of experience in an industrial setting with proven results is a strong plus
- Knowledge of queuing theory
- Familiarity with scheduling algorithms and heuristics
- Basic statistics for parameter estimation and data analysis
- Simulation both Discrete Event and Monte Carlo (desired)
- Programming skills enough to implement your own algorithms/ideas
- Ability to work in a UNIX/Linux environment
- Ability to write scripts using one of the common scripting languages (Perl, Python, etc.)
- Excellent written and oral communication skills with strong inter-personal interaction skills
- Facilitation and presentation skills.
- Leadership and teamwork skills
EOE
Interested in talking with us? Please apply directly at NVIDIA.COM