Senior AI Infrastructure SRE - GPU Clusters & MLOps

ByteDance; San Jose (CA); On-site; USD 180,000 - 280,000; Full time

Job summary

ByteDance’s Seed Infra SRE team manages the compute infrastructure for foundation models, including distributed training, heterogenous hardware, and high-speed networks. You will own training/inference stability, optimize resource scheduling, and lead deployment of large-scale GPU/NPU clusters while ensuring data reliability and cost-efficient capacity planning.

You will partner with product, algorith m, and operations teams to implement robust MLOps practices, drive performance optimizations,

Qualifications

• Bachelor's degree in CS or related field; 3+ years in large-scale ML/MLOps infra or similar work.
• Proficient in Python and/or Go; strong Linux and shell scripting; automation tooling experience.
• Expert in containerization and orchestration (Docker/Kubernetes); hands-on with large clusters.
• Familiar with high-speed networking (RDMA) and high-performance storage.
• Strong scheduling experience and ability to optimize compute utilization.

Responsibilities

• Own stability of training and hetero inference systems for large-scale AI workloads; optimize performance and resolve bottlenecks.
• Design and iterate resource scheduling systems; improve utilization and explainability; support multi-type workloads.
• Lead deployment and operation of large-scale networks and heterogeneous compute clusters (GPU/NPU); ensure topology and bandwidth.
• Own architecture and operations of high-performance storage; ensure I/O performance and data reliability.
• Drive capacity planning,Compute pool build-out, and cost control for scalable delivery.
• Build monitoring, alerting, and automated remediation across training and inference paths; rapid issue resolution.
• Collaborate with algorithm/product/ops teams to deliver MLOps solutions and raise technical bar.

Skills

Python

Go

Linux

Shell scripting

Distributed systems

Education

Bachelor's degree in CS/related field

Tools

Docker

Kubernetes

Kata containers

RDMA

Job description

ByteDance’s Seed Infra SRE team manages the compute infrastructure for foundation models, including distributed training, heterogenous hardware, and high-speed networks. You will own training/inference stability, optimize resource scheduling, and lead deployment of large-scale GPU/NPU clusters while ensuring data reliability and cost-efficient capacity planning.

You will partner with product, algorith m, and operations teams to implement robust MLOps practices, drive performance optimizations,