Position OverviewWe are seeking a Staff AI Observability & Telemetry Engineer to architect the "nervous system" of our AI-native NeoCloud platform. This role goes beyond standard monitoring; you are responsible for building the high-fidelity perception layer required to orchestrate massive-scale AI infrastructure. You will capture, store, and make sense of millions of hardware and software signals per second, enabling our SREs, automated remediation agents, and external customers to peer deep into the performance of their GPU workloads and the underlying network fabric. You will define the telemetry standards that drive our autonomous operations, ensuring we can detect, diagnose, and resolve hardware and software bottlenecks in real-time.
Key Responsibilities- Architect and scale a high-cardinality telemetry infrastructure using highly available time-series databases (e.g., VictoriaMetrics, Thanos, or Mimir) capable of handling massive ingestion rates.
- Integrate complex hardware-level exporters (NVIDIA DCGM, network switch telemetry, IPMI/Redfish) directly into the Kubernetes observability stack to provide a unified view of the cluster.
- Build eBPF-based diagnostic tools to trace network congestion, kernel-level I/O latency, and distributed training bottlenecks across the cluster.
- Develop automated dashboards and alerting pipelines that trigger proactive cordoning of degraded hardware before it impacts customer training jobs.
- Design the metric pipelines required for accurate, multi-tenant consumption billing based on real-time GPU and network utilization metrics.
- Collaborate with the GPU Systems and Scheduling teams to create observability standards for "AI-native" workloads, ensuring deep insight into job efficiency and resource utilization.
- Lead technical design reviews for observability architecture, mentoring team members on best practices for high-performance telemetry collection and analysis.
Qualifications- Bachelor's or Master's degree in Computer Science, Electrical Engineering, or a related field.
- 6+ years of software or site reliability engineering, with deep, hands-on expertise in the Prometheus/OpenTelemetry ecosystem.
- Advanced proficiency in Go and extensive experience writing custom Kubernetes metric exporters and operators.
- Hands-on experience with kernel-level tracing tools (eBPF, BCC) and deep performance tuning of Linux systems.
- Strong familiarity with AI hardware metrics (GPU power states, SM utilization, memory bandwidth) and high-performance network telemetry.
- Proven track record of operating, debugging, and scaling large-scale telemetry stacks in high-performance computing or cloud environments.
- Strong technical leadership skills; ability to influence architectural decisions and align cross-functional teams around observability standards.
- Excellent communication skills, with the ability to translate complex system requirements into manageable engineering milestones.
- Experience working in high-velocity, high-growth engineering environments is strongly preferred.
