Staff Software Engineer, GPU Inference
Cerebras · Toronto, CAN · Sunnyvale, CA
CHECKED 8H AGO
Infrastructureremotestaff
At a glance
Highlights
- Build disaggregated inference pairing GPU prefill with wafer-scale decode.
- Own reliability and performance of rack-scale AI inference in production.
- Cut time to first token and boost throughput for frontier models.
Ideal candidate
Veteran systems owner who loves deep debugging across hardware and software and turning one-off investigations into reliable, high-performance production inference.
Requirements
- 8+ years of software engineering experience with ownership of complex production systems
- Experience building, operating, or optimizing production inference for LLMs or demanding GPU workloads
- Strong C++ and Python skills with multithreading, concurrency, and memory management
- Experience optimizing Mixture-of-Experts or multimodal models
- Knowledge of GPU kernel optimization, operator fusion, attention kernels, and communication-computation overlap
- Experience with reduced-precision inference including BF16, FP8, FP4, INT8, or INT4
- Experience building numerical-comparison, determinism, or performance-regression test systems
Lab profile
About Cerebras
Builds wafer-scale AI chips and cloud systems for ultra-fast training and inference.
What they build
- Wafer-scale AI chips
- AI training systems
- Cloud inference service
Known for
- 56x larger chips than GPUs
- 10x faster inference than GPU clouds
- Industry-leading training speed