About the job
We’re looking for an experienced machine learning researcher / engineer who can help us push the frontiers of agentic LLM systems. As a part of the team, you will help drive exploration and development of agentic techniques and have the opportunity to build the models that power our agentic solutions. Agentic LLM systems are being deployed widely across enterprise companies including through Cohere’s North platform. In this role, you’ll be working with a team developing new strategies for training models for advanced agent capabilities including reasoning, tool use, and memory. This includes developing data-generation techniques for post-training (SFT and RL*) Cohere’s models. Model advancements have direct impacts on North and other Cohere products creating an exciting opportunity where core model development leads to direct product advancements.
Responsibilities
Design and develop novel agentic solutions
Improve upon SOTA on hard agentic tasks
Research the next-generation of on-line learning-from-experience self-improvement
Work with partner teams (Reasoning, Post-training, Pre-training, etc.) to improve performance of agentic system
Work with an amazing team of researchers and engineers pushing the boundaries
Qualifications
Minimum
Have a PhD in computer science or related field or similar industry research experience
Strong software engineering skills
Proficiency in Python and experience with ML-related code (e.g., pytorch, numpy, etc.)
Experience with LLMs and agentic frameworks
Experience with post-training LLMs (SFT, PEFT, or RL*)
Experience with building synthetic data generation pipelines
Preferred
No preferred qualifications listed.