Teams at Sciforium

The role involves optimizing and maintaining the software stack for large-scale AI training, focusing on CUDA/ROCm, JAX, and PyTorch to ensure efficiency and stability in distributed training systems.

Design, implement, and optimize custom GPU kernels (CUDA/ROCm, PTX, Triton/JAX Pallas) for large-scale LLM training and inference; profile performance, integrate kernels into PyTorch/JAX and internal runtimes, collaborate with researchers and hardware vendors, and build tooling, benchmarks, and testing for reproducible performance.

Develop the core systems for multimodal AI models, focusing on building model serving platforms and enhancing developer experience with tools and APIs.