Principal Performance Architect - Ontap

Job Summary

As a Performance Architect, you will lead the ONTAP performance team for the various hardware and software projects for ONTAP releases. You will work closely with architects across various engineering development teams to drive performance agenda, establish performance direction, guide performance team’s technical work across all projects. You will have the opportunity to directly impact the roadmaps of all hardware and software groups. You will be able to drive the analysis, benchmarking, and optimization of emerging workloads to capitalize on industry trends including Generative and Agentic AI, and lead ONTAP’s AI benchmark strategy, including industry-standard submissions such as MLPerf Storage. This role is within the Research and Development function whose overall focus is on technology advances, competitive market and customer workload requirements, product cost and time-to-market, and product quality.

Job Responsibilities

Evaluate and propose key architecture options to drive performance
Evaluate ONTAP competitiveness and drive differentiation with a performance focus, including against competitors’ published AI benchmark results
Optimize ONTAP performance for Generative and Agentic AI workloads across the AI data pipeline: data ingest and preparation, training, checkpointing, inference, KV-cache, and RAG/vector search
Own ONTAP’s AI benchmark program: plan, run, and publish industry benchmark submissions (e.g., MLPerf Storage, SPECstorage AI_IMAGE) that show competitive leadership
Define AI benchmark methods and workload models (e.g., checkpoint/restore patterns, inference traces) that turn real customer AI pipelines into tests anyone can repeat
Work with GPU and AI-ecosystem partners on reference architectures, certifications, and high-performance data paths (GPUDirect Storage, RDMA, NFS over RDMA, pNFS)
Work closely with engineering and product management to drive the performance agenda.
Evaluate design alternatives for enhanced performance, and prototype opportunities for performance enhancements.
Design, implement, execute, analyze, interpret, socialize and apply storage-oriented performance measurements and their results, including the creation of tools and automation for running AI benchmarks at GPU-cluster scale
Turn benchmark results into technical reports, competitive briefs, and conference talks for product marketing and field teams

Job Requirements

In-depth knowledge of technology trends, architecture differentiation, competitive positioning, and workload optimizations.
Hands-on experience with AI/ML benchmarks and frameworks and an understanding of how AI training and inference stress storage (throughput, metadata, small random reads, checkpoint write bursts)
Familiarity with GPU server architectures and high-performance data paths: RDMA, NVMe-oF, GPUDirect Storage, parallel file access
Extensive knowledge and experience in computer operating systems, computer architecture and design, data structures, software design and programming practices; systems programming in C and/or C++
Strong scripting skills - Perl, Python (primarily within JupyterHub), and Shell
Knowledge of storage systems and file systems highly desirable
Experience publishing benchmark results or taking part in industry groups (e.g., MLCommons, SPEC) a plus
Exceptional presentation and influencing skills.
The ability to make accurate work estimates and develop predictable plans.
Demonstrated willingness to take on additional tasks and responsibilities that positively contribute to realization of team, department, and company goals.

Education

Typically requires a minimum of 12-15 years of relevant experience.
A Bachelor of Science, Master’s of Science, or PhD Degree in Electrical Engineering or Computer Science; or equivalent experience is required.