CUDA C++ engineers can design GPU kernels at Baseten, optimizing matrix multiplication and attention for production machine learning inference. Work model is not specified.

As a GPU Kernel Engineer, you will support Baseten’s Model Performance team and its inference stack. You will write CUDA and PTX code, apply GPU architecture knowledge, and optimize contemporary AI workloads. The work includes collaboration with research teams and production systems.

Key Responsibilities:

- Design high-performance kernels for matrix multiplications
- Optimize attention mechanisms and mixture-of-experts routing
- Apply memory coalescing and warp-level programming
- Use tensor cores and compute-memory overlap
- Contribute to internal and open-source GPU libraries

Requirements:

- Understand GPU architecture and programming paradigms
- Know memory hierarchy and thread organization
- Mitigate race conditions using synchronization techniques
- Demonstrate C++ proficiency and profiling-tool experience
- Understand CUDA C++ APIs and bandwidth optimization

Build Baseten’s inference performance through CUDA C++, GPU architecture, kernel optimization, and production-focused machine learning systems.

📌 GPU Kernel Engineer CUDA C++ (Montreal)
🏢 Baseten
📍 Montreal