SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Cohere is a leading enterprise AI company building cutting-edge foundation models and end-to-end products for businesses. As a Member of Technical Staff in Post-Training, you will design and execute large-scale reinforcement learning experiments to train large language models (LLMs) and develop methods that enable them to generalize beyond their training distribution.
You will take ownership of multiple dimensions of post-training research and development. Your core responsibilities include designing, running, and analyzing large-scale RL post-training experiments such as RLHF and RL from verifiable rewards. You will research and develop methods to improve out-of-distribution generalization through reward design, environment and task diversity, curricula, and regularization techniques. You will build held-out evaluations that measure generalization to unseen tasks, domains, and formats while catching reward hacking and benchmark overfitting. You will diagnose training dynamics and failure modes including instability, reward over-optimization, and capability regressions. Finally, you will translate successful research into robust, reproducible training recipes that power production models.
You will collaborate closely with post-training data infrastructure, evaluation, and pre-training teams. Cohere is a global company with offices in Toronto, London, New York City, San Francisco, Montreal, Paris, Berlin, and Seoul, and this role offers full flexibility to work remotely or from any office location with no minimum in-office requirement.
QUALIFICATIONS:
- Strong research and software engineering skills with hands-on experience training LLMs
- Experience applying reinforcement learning to LLMs (e.g., PPO, GRPO, RLHF, RLVR), ideally at scale
- Track record of work on generalization, robustness, or distribution shift, demonstrated through publications or shipped models
- Rigorous experimental practice: clean ablations, careful controls, and sound conclusions from noisy results
- Experience with distributed training (e.g., JAX, PyTorch, FSDP, Megatron) is a plus
- Excellent communication skills
- Deep understanding of data quality principles and best practices; willingness to dive deep into data with strong intuition for assessing dataset quality and anticipating how data characteristics impact model training outcomes
About Cohere
AI / Data / Infrastructure — enterprise generative AI models and tooling for businesses.
