All interviews
Cohere logo

Cohere

Senior

Build data generation, post-training, and evaluation methods to make Cohere's agentic LLMs safer and more secure

Cohere's Safety for Agents team is hiring a senior MTS to work on data generation, post-training algorithms, and evaluation methods for LLMs that can access external tools and take actions in the world. The role blends ML research, experimental design, and hands-on engineering with human-annotator data pipelines, requiring publication-level research credibility (a paper at a top venue like NeurIPS/ICML/ACL) plus strong software engineering. Expect deep technical questions on distributed LLM training, dataset quality/bias analysis, and designing evaluations for agent safety and robustness.

Step into this interview

Free · a live voice mock calibrated to this exact role

Practice this interview

What this interview tests

  • Post-training algorithms and evaluation methods for agentic LLM safety
  • Designing and running data collection tasks with human annotators
  • Dataset quality, bias, and suitability analysis for training data
  • Hands-on distributed LLM training experience
  • Statistical rigor in evaluating scientific/ML experiments
  • Reasoning about agent-specific failure modes vs. static-generation safety

Common question themes

Describe an experiment you designed to evaluate data collection or model performance, and how you validated it statistically

How would you design an annotation task and catch bias or low-quality labels in the resulting dataset

Walk through your hands-on experience training an LLM on distributed infrastructure and a bottleneck you solved

Tell me about owning an ambiguous safety/research problem from question to shipped evaluation

How do agent-specific safety failure modes differ from standard LLM safety issues, and how would you evaluate for them

Discuss a publication or research contribution and how it applies to industrial safety work

View the original posting

Related interviews