Staff GPU Inference Engineer — Production AI Acceleration

Cerebras; Sunnyvale (CA); On-site; USD 250,000 - 360,000; Full time

Benefits offered by this job

Breakthrough AI platform

Open source research

Work on one of the fastest AI superwec

Startup vitality

Non-corporate culture

Job summary

Cerebras Systems seeks a senior software engineer to productionize and optimize its GPU-serving stack across custom APIs, vLLM, and ROCm. You will work on AI inference systems, ensuring numerical correctness, observability, and high performance at rack scale in Sunnyvale.

The role requires deep debugging across application, runtime, distributed systems, and hardware layers, with a focus on time-to-first-token, throughput, and capacity efficiency in production.

Qualifications

• 8+ years of software engineering experience in complex production systems.
• Experience building/operating/optimizing production inference systems for large language models or similar GPU workloads.
• Strong programming in C++ and Python, with multithreading and memory management.

Responsibilities

• Productionize the GPU inference stack across API services, model-serving workers, vLLM, PyTorch, ROCm, GPU nodes, networking, and rack-scale infrastructure.
• Own GPU operational readiness including deployment, upgrades, health checks, and capacity management.
• Drive reliability with service-level indicators and automated recovery.
• Improve inference performance: time to first token, throughput, tail latency, and GPU utilization.
• Debug across application, runtime, hardware, and distributed systems layers to ensure numerical correctness and robustness.

Education

Bachelor’s degree in Computer Science, Computer Engineering, Electrical Engineering, or a related discipline

Tools

vLLM

SGLang

TensorRT-LLM

Triton Inference Server

PyTorch

ROCm/HIP

Job description

Cerebras Systems seeks a senior software engineer to productionize and optimize its GPU-serving stack across custom APIs, vLLM, and ROCm. You will work on AI inference systems, ensuring numerical correctness, observability, and high performance at rack scale in Sunnyvale.

The role requires deep debugging across application, runtime, distributed systems, and hardware layers, with a focus on time-to-first-token, throughput, and capacity efficiency in production.