Principal Software Engineer, At-Scale Reliability and Fleet Intelligence — CSP Engagements

What You'll Be Doing

• Drive reliability work streams with CSP engineering teams — ensuring shared understanding of MTBI measurement methodology, failure classification, and health monitoring architecture
• Gather and synthesize CSP fleet reliability data — identify failure patterns that appear across multiple customers and champion improvements back into NVIDIA's firmware, driver, and hardware teams
• Define consistent MTBI measurement methodology that works across different CSP monitoring environments and operational practices
• Conduct fleet-scale failure pattern analysis using statistical methods (Pareto, survival analysis, Weibull) to classify failures as systemic, environmental, or configuration-specific
• Drive fleet health monitoring integration architecture — ensure NVIDIA's health agents, telemetry, and reporting align with CSP operational workflows and automation
• Define burn-in reliability test environment and cluster certification criteria in collaboration with quality teams, validating with customers that criteria are meaningful
• Collaborate with CSPs to ensure reliability-related integration work (health monitoring deployment, telemetry pipeline, alerting configuration) is complete ahead of at-scale launch
• Develop predictive failure models using fleet telemetry and validate their effectiveness in customer environments

What We Need To See

• 15+ years of experience in systems software at datacenter scale, or reliability engineering with focus on at-scale challenges
• BS or MS in Computer Science, Electrical Engineering, Statistics, or related field (or equivalent experience)
• Deep expertise in multi-NUMA, rack-scale system software and firmware. Statistical failure analysis methods: MTBF/MTBI calculation, Pareto analysis, root cause classification
• Experience with fleet-level telemetry and observability systems: time-series databases, anomaly detection, health scoring, event correlation
• Understanding of hardware failure modes in large-scale GPU/accelerator deployments — ability to classify and prioritize across compute, interconnect, memory, power, and thermal domains
• Experience defining or operating burn-in, stress testing, or certification frameworks for complex hardware systems. Familiarity with predictive maintenance or anomaly detection approaches applied to fleet health data
• Customer obsession — genuine passion for understanding fleet reliability challenges at scale and translating them into actionable engineering priorities
• Strong communication — ability to present statistical reliability findings to both deep technical audiences and executive leadership. Demonstrated success driving cross-functional improvements across hardware, firmware, and software teams without direct authority

Ways To Stand Out From The Crowd

• Experience in fleet reliability at a hyperscaler (hardware health, fleet reliability at leading CSP/Hyperscaler)
• Familiarity with NVIDIA GPU error taxonomy (Xid errors, NVLink error counters, thermal events, CPER records)
• Experience building health scoring or predictive failure models for accelerator or HPC infrastructure
• Background in defining MTBI/MTBF measurement standards or certification programs for complex multi-component systems
• Understanding of how reliability data flows from device firmware through telemetry pipelines to fleet-level dashboards and automated remediation