Graduate Intern – Power Data Analysis for AI Data Centers
About The Position
The Digital Twins and Scalable Workflows group in NLR’s Computational Science Center seeks a graduate student researcher to analyze data center power, particularly the electrical behavior of AI workloads. The project measures the electrical power of GPU racks on NLR’s Kestrel supercomputer as AI training, inference, and agentic workloads run. Those measurements will be used to develop models that predict realistic power behavior for workloads that have not been measured. The researcher will review experiment measurements, identify anomalies and trace their causes using system telemetry and profiling data, and label power ramps and other critical events. They will also build an automated event detector trained and evaluated on labeled data. The successful candidate will work with NLR researchers in high-performance computing, power systems, and data center operations, and may present and publish their work.
Requirements
- Minimum cumulative GPA of 3.0.
- Must meet the applicable educational requirements before the employment start date.
- Graduate student eligibility: enrolled full-time in a master’s degree program at an accredited institution; or have earned a master’s degree within the past 12 months; or have completed a master’s degree and be enrolled as a PhD student at an accredited institution.
- Applicants are responsible for uploading official or unofficial school transcripts as part of the application process.
- If selected, a letter of recommendation is required as part of the hiring process.
Nice To Haves
- Careful attention to detail and ability to follow and help refine a structured review process.
- Excellent written and verbal communication skills, including clear documentation of findings.
- Ability to work independently and raise questions early.
- Experience analyzing time-series or signal data, such as spectral analysis or event detection.
- Familiarity with GPU computing, AI/ML workloads, or HPC systems such as Slurm.
- Programming experience in Python for data analysis.
- Experience with machine learning for classification or detection.
- Background in power systems or electrical measurements.
- Experience with data labeling, annotation, or data quality review.
Responsibilities
- Review power measurements from AI workload experiments for anomalies and data quality issues using an interactive visualization workbench.
- Hand-label power ramps, load steps, and other critical events, and identify regions exemplifying characteristic workload behavior according to a documented labeling protocol.
- Investigate unexplained behavior by correlating power measurements with GPU telemetry, profiling data, job logs, and system configuration, and document findings.
- Design, build, and evaluate an automated detector for power ramps and anomalies, trained and tested on labeled data.
- Present results and contribute to technical reports, papers, and conference presentations.
Benefits
- Medical insurance
- Dental insurance
- Vision insurance
- 403(b) Employee Savings Plan with employer match
- Sick leave where required by law
- Potential eligibility for performance-, merit-, and achievement-based monetary awards
- Some positions may be eligible for relocation expense reimbursement
- Professional development opportunities
Stand Out From the Crowd
Upload your resume and get instant feedback on how well it matches this job.
What This Job Offers
Career Level
Intern
