Senior Member of Technical Staff, Safety and Security for Agents

Role overview

Cohere is a leading enterprise AI company that builds foundation models and end-to-end products for real-world business applications. As a Senior Member of Technical Staff in the Safety and Security for Agents team, you will focus on developing safer, fairer, and more secure large language models. Your work will concentrate on data generation, post-training algorithms, and evaluation methods for the next generation of models capable of accessing external resources and taking autonomous actions.

Key responsibilities

Apply strong statistical skills to evaluate scientific experiments related to data collection methodologies and model performance outcomes.
Develop and implement software engineering solutions to test and validate approaches to safety challenges in language models.
Design and oversee data collection tasks, including coordination with human annotation teams to gather training data.
Analyze datasets to assess quality, identify biases, and determine their suitability for training machine learning models.
Work hands-on with large language model training on distributed computing infrastructure.
Evaluate and enhance the generalizability and robustness of machine learning systems.
Collaborate with cross-functional teams including researchers, engineers, product specialists, and policy experts to communicate findings and drive improvements.

Requirements

Demonstrated expertise in Python programming and machine learning frameworks such as PyTorch, TensorFlow, or JAX.
Strong background in designing and conducting data collection initiatives with human annotators.
Proven ability to analyze and assess dataset quality, biases, and suitability for machine learning applications.
Hands-on experience training large language models on distributed training systems.
Publication record with one or more papers accepted at top-tier academic venues including NeurIPS, ICML, ICLR, AIStats, MLSys, JMLR, AAAI, Nature, COLING, ACL, or EMNLP.
Excellent communication skills for cross-functional collaboration and presenting technical findings.

What makes this role worth considering

You will join a small, autonomous team with significant decision-making authority responsible for shaping the safety and trustworthiness of next-generation language models. The existing team spans multiple locations including London, Edinburgh, Paris, Toronto, and New York, with full remote-friendly flexibility as long as you maintain approximately 50% working day overlap with UK/EU timezone.