Member of Technical Staff - Post-Training (Reinforcement Learning & Generalization)

O texto do emprego está no idioma original.

Cohere, the security-first enterprise AI company behind the Command models, is hiring a Member of Technical Staff in Post-Training to train its large language models with reinforcement learning and make them generalize beyond their training distribution. You'll design and run large-scale RL post-training experiments (RLHF, RL from verifiable rewards), research methods such as reward design, environment and task diversity, curricula and regularization, build held-out evaluations that show whether a model has really learned a skill, and turn what works into the recipes behind Cohere's production models, working with the post-training data, evaluation and pre-training teams. The role can be fully remote from Europe or based at a Cohere office, with no minimum in-office requirement; the posting also lists New York and Toronto. Pay CA$295,000–CA$535,000 plus equity, with multiple ranges by location. Full-time. Verified live on Cohere's Ashby board.