Cohere

Research internship focused on combining self-distillation and reinforcement learning for LLMs, with applications to code generation and agentic tasks. Investigates mechanisms for handling extremely large rollouts in RLVR to advance state-of-the-art in LLM training.

Cohere Labs is seeking Research Engineers to build experiments, debug models, scale training pipelines, and turn research ideas into working systems. This role involves implementing new methods, running large-scale experiments, and shaping research infrastructure, with a focus on deep learning frameworks, model optimization, and collaboration with scientists.