Compute Server Platform Architect
Posted Updated 3
views
Summary
Own server-side platform architecture for Cerebras AI clusters: define server configurations and baselines (CPU, memory, PCIe, NIC, NVMe, BIOS/firmware), build capacity and performance models, benchmark and qualify vendor platforms, and support deployment debugging. Core stack is x86 servers, Linux, RDMA/RoCE networking, and C/C++/Python.
You will own server-side platform architecture for AI clusters. You will define server configurations, capacity formulas, CPU, memory, PCIe, networking, storage, and firmware baselines. You will model and benchmark performance, lead vendor engagements, establish qualification criteria, and troubleshoot deployment regressions.
Responsibilities
- Own architecture for cluster server roles, configurations, and lifecycle strategy
- Define server formulas, capacity planning, and headroom policy
- Specify CPU, memory, PCIe, NIC, and NVMe platform configurations
- Translate runtime flows into hardware requirements
- Develop and validate performance and scaling models
- Define operating system, BIOS, firmware, and driver baselines
- Evaluate emerging server technologies
- Lead vendor engagements and qualification efforts
- Support deployment debugging and root-cause analysis
Requirements
- PhD and 8+ years of industry experience, or BS or MS and 10+ years of industry experience
- 5+ years of server platform architecture, systems performance engineering, or large-scale infrastructure design
- x86 server architecture
- Linux systems
- NIC behavior
- RDMA
- RoCE
- NVMe
- Capacity and performance modeling
- Benchmarking
- Vendor platform evaluation
- C
- C++
- Python