Member of Technical Staff - Post-Training (Reinforcement Learning & Generalization)
CA$295k - CA$535k • Remote • Europe • FullTime
Posted 7h ago
About the job
Cohere is a leading enterprise AI company focused on building cutting-edge foundation AI models and end-to-end products for real-world business problems. We are training and deploying frontier models for enterprises building AI systems, aiming to drive widespread AI adoption. This role is for a Member of Technical Staff in Post-Training, responsible for training large language models (LLMs) using reinforcement learning and developing methods for generalization beyond training distributions. You will design training recipes, environments, and evaluations to ensure models truly learn skills and translate successful research into production-ready recipes. Collaboration with post-training data infrastructure, evaluation, and pre-training teams is essential.
Responsibilities
- Design, execute, and analyze large-scale RL post-training experiments, including RLHF and RL from verifiable rewards.
- Research and develop methods to enhance out-of-distribution generalization through reward design, environment/task diversity, curricula, and regularization.
- Build held-out evaluations to measure generalization to unseen tasks, domains, and formats, and to detect reward hacking and benchmark overfitting.
- Diagnose training dynamics and failure modes such as instability, reward over-optimization, and capability regressions.
- Translate successful research findings into robust and reproducible training recipes for production models.
Requirements
- Strong research and software engineering skills with hands-on LLM training experience.
- Experience applying reinforcement learning to LLMs (e.g., PPO, GRPO, RLHF, RLVR), preferably at scale.
- Demonstrated work on generalization, robustness, or distribution shift via publications or shipped models.
- Proficiency in rigorous experimental practices, including clean ablations, careful controls, and sound conclusions from noisy results.
- Excellent communication skills.
- Deep understanding of data quality principles and best practices, with strong intuition for assessing dataset quality and its impact on model training.
Benefits
- Weekly lunch stipend of $75/£75 or equivalent.
- Full health and dental benefits, including a mental health budget.
- RRSP matching, 401K, Pension Scheme.
- 100% Parental Leave top-up for up to 6 months.
- Annual enrichment benefits (Arts & culture, fitness/wellness, quality time, workspace improvement).
- Education & learning stipend for conferences, courses, and coaching.
- 6 weeks (30 working days) of paid vacation.
- Budget for traveling to other offices for remote employees.
- Annual company offsite.
- Co-working benefit for those not near an office.
- Home office stipend of $500.