jobs

Member of Technical Staff, Training Performance Engineer

Who are we?Cohere is the leading security-first enterprise AI company. We build cutting-edge foundation AI models and end-to-end products that are designed to solve real-world business problems.We’re training and deploying frontier models f

Company

Cohere

Location

London; New York; Paris; Toronto; Montreal

Status

Open

Posted

2025-02-20T10:45:59.057+00:00

A performance engineering role on Cohere's Pre-Training team, focused on making large language model training faster and more efficient. The work covers writing scalable training software, building low-level CUDA and Triton kernels to maximise accelerator utilisation, and studying how architectural choices affect training throughput and quality. The posting asks for strong software engineering, Python with JAX, PyTorch or XLA/MLIR, GPU kernel experience, large-scale distributed training, and familiarity with Transformer-style autoregressive models. Publication at venues such as NeurIPS or ICML is listed as a bonus. The team works mainly between ET and CET.