岗位职责
The Server Management DevOps team is responsible for the end-to-end lifecycle management of servers across ByteDance’s self-built data centers in the United States and Europe. Our scope covers new hardware introduction, data center delivery, production operations, hardware maintenance, configuration and firmware changes, capacity migration, asset decommissioning, data sanitization, and hardware reuse. The team serves as a central engineering and coordination point across multiple functions, including: - Hardware New Product Introduction (NPI) - Server and data center operations - Field maintenance and infrastructure management - Hardware vendors and service providers - Supply chain and asset management - Infrastructure platform and automation engineering teams As ByteDance continues to expand its AI infrastructure, the team is taking on an increasingly important role in introducing, productionizing, and operating high-density GPU platforms at scale. About the Role We are looking for an experienced and hands-on Senior Production Systems Engineer with deep expertise in large-scale GPU infrastructure, Linux systems, hardware lifecycle management, automation, and production operations. In this role, you will lead the introduction and productionization of current- and next-generation AI infrastructure, including rack-scale and high-density GPU platforms such as NVIDIA GB200/GB300 NVL72, HGX or DGX B200/B300, Vera Rubin NVL72, and comparable accelerator systems. You will own critical work across platform evaluation, system and firmware integration, data center readiness, deployment validation, fleet onboarding, monitoring, incident response, and long-term operational reliability. You will also work directly with AI training and inference environments to ensure that the underlying infrastructure meets real workload requirements. This is a senior individual-contributor role requiring strong technical judgment, hands-on engineering ability, and the capacity to lead complex global infrastructure initiatives across organizational boundaries. Key Responsibilities - Advanced GPU Platform Introduction: Lead the evaluation, qualification, integration, and production rollout of next-generation GPU platforms, including GB200/GB300 NVL72, B200/B300 systems, Vera Rubin, and future rack-scale AI infrastructure. - End-to-End Production Readiness: Define launch criteria and readiness plans spanning server hardware, firmware, BMC, operating systems, drivers, GPU software stacks, networking, storage, security, telemetry, and operational tooling. - Rack-Scale Integration and Fleet Operations: Partner across hardware, data center, network, storage, power, cooling, and vendor teams to resolve system-level challenges and improve GPU fleet availability, utilization, serviceability, and lifecycle management. - Systems, Performance, and Reliability Engineering: Diagnose complex Linux, hardware, firmware, PCIe, NVLink/NVSwitch, network, storage, and memory issues, while developing qualification, burn-in, benchmarking, health-check, and regression-testing frameworks. Improve the availability, utilization, serviceability, and lifecycle management of large GPU fleets across multiple data center regions. - Automation, Observability, and AI-Assisted Operations: Build scalable automation and actionable telemetry for provisioning, configuration, monitoring, fault detection, remediation, repair, and lifecycle operations; apply AI technologies to incident triage, troubleshooting, knowledge retrieval, and automated remediation. - Technical and Cross-Functional Leadership: Establish engineering standards, operational procedures, and long-term support models; mentor engineers and drive complex infrastructure programs across internal teams, supply-chain partners, and external vendors. - On-Call and Global Operations: Participate in a global on-call rotation and provide senior-level leadership during critical production incidents. Occasional travel to data centers, integration facilities, or vendor sites may be required. Lead the investigation of complex and high-impact production incidents, coordinate mitigation across teams and vendors, perform root-cause analysis, and ensure that preventive actions are implemented and measured.
职位要求
Minimum Qualifications - Bachelor’s degree in Computer Science, Computer Engineering, Electrical Engineering, Information Technology, or a related field, or equivalent practical experience, with 5+ years of experience in production systems, infrastructure engineering, Site Reliability Engineering, DevOps, hardware systems engineering, or large-scale data center operations. - Proven hands-on experience introducing and productionizing large-scale GPU infrastructure, including ownership of hardware NPI or fleet onboarding across qualification, system integration, deployment, production validation, operational handoff, and post-launch reliability. Experience must include a recent-generation platform such as NVIDIA GB200/GB300 NVL72, HGX or DGX B200/B300, or a technically comparable rack-scale AI system. - Deep knowledge of Linux administration and troubleshooting, combined with a strong understanding of server architecture and management technologies such as kernels, drivers, BIOS/UEFI, BMC, Redfish, firmware, PCIe, NVMe, NICs, DPUs, hardware telemetry, and failure diagnostics. - Hands-on experience deploying or supporting distributed AI training or inference workloads using containerized and orchestrated environments, with working knowledge of technologies such as CUDA, GPU drivers, NCCL, NVLink, NVSwitch, RDMA, InfiniBand, RoCE, or high-performance Ethernet. Familiarity with high-density data center requirements, including liquid cooling, power delivery, rack integration, structured cabling, network fabrics, and deployment safety. - Proficiency in Python, Go, Bash, or another programming language for production-grade infrastructure automation, along with practical experience applying AI or large language models to engineering workflows such as incident analysis, troubleshooting, knowledge retrieval, code generation, or automated remediation. - Experience building and operating monitoring, telemetry, hardware management, or automated remediation platforms at substantial scale, with measurable results in fleet availability, deployment efficiency, incident reduction, operational efficiency, or reliability. - Strong systems-thinking, technical leadership, and communication skills, with the ability to troubleshoot across hardware and software layers and lead complex initiatives involving global teams, infrastructure partners, and vendors. Professional proficiency in English is required. Preferred Qualifications - Direct experience deploying and operating GB200 or GB300 NVL72 systems, including compute and NVLink switch trays, Grace CPUs, Blackwell GPUs, ConnectX networking, BlueField DPUs, rack management, firmware dependencies, power delivery, and liquid-cooled data center integration. - Experience planning, qualifying, or preparing production and operational environments for next-generation platforms such as NVIDIA Vera Rubin NVL72. - Experience operating large-scale GPU clusters across multiple data centers or geographic regions. - Strong understanding of distributed AI workload behavior and performance analysis, including collective communication, multi-node training, inference serving, GPU scheduling, checkpointing, workload-related bottlenecks, DCGM, NCCL testing, CUDA profiling, and network-fabric telemetry. - Experience with Kubernetes GPU Operator, Slurm, Ansible, configuration management, infrastructure as code, CI/CD, or large-scale provisioning and orchestration systems. - Experience working directly with OEMs, ODMs, component suppliers, or GPU platform vendors throughout qualification, technical escalation, root-cause analysis, and corrective-action processes. - Contributions to infrastructure engineering communities, technical publications, patents, open-source projects, or relevant industry standards.
