MLPerf Software Engineer
MLCommons is a global non-profit AI engineering consortium that develops open benchmarks, datasets, data standards, and related tooling for measuring AI performance, quality, safety, and risk.
Funding history
Investors
About MLCommons
MLCommons collaborates with industry, academia, non-profits, and other members to improve AI systems through reproducible benchmark suites, safety evaluation, open datasets, and data standards. Its work includes MLPerf benchmarks, AILuminate AI safety benchmarks, Croissant dataset metadata, and MLCube portability conventions.
Skills
About the Role
You will design and build the systems, infrastructure, and automation that power MLPerf benchmarks. You will own technical decisions from architecture through implementation, improve existing pipelines, resolve complex issues, and document and present technical designs and outcomes.
Requirements
- Experience designing systems and infrastructure, including architecture decisions, trade-off analysis, scalability, and reliability.
- 4+ years of post-academic software engineering experience or equivalent open-source or industry impact.
- Python.
- Bash.
- GitHub Actions or equivalent CI/CD.
- Artifact and version management.
- Docker containers.
- ML concepts, including inference and training basics.
- Performance benchmarking or optimization.
Responsibilities
- Architect and design benchmarking systems, infrastructure, and tooling.
- Design scalable automation for benchmark submission, validation, review, publication, and reproducibility.
- Operate and improve existing MLPerf pipelines.
- Translate roadmap goals into technical designs and deliver them.
- Collaborate on end-to-end benchmarking pipelines.
- Debug and resolve issues across automation, submissions, and benchmark harnesses.
- Produce design documentation.
- Present technical designs, progress, and outcomes to working groups.
Benefits
- Fully remote work.
- Flexible working environment.
Hiring Process
Apply by emailing jobs@mlcommons.org with your resume and relationships for three professional references; include the job title in the subject line. References will not be contacted without permission.