Build pretraining data systems at Cohere as a Machine Learning Engineer, using Python and data pipelines in a remote-friendly setting. This Machine Learning Engineer role focuses on pretraining data for Cohere’s advanced language models. You will combine research and engineering across data ablations, data mixtures, data modeling, and curation. Your work supports improvements in training throughput and accelerator utilization while advancing natural language processing. Key Responsibilities:
Conduct data ablations to assess quality and test data mixtures Develop data modeling techniques for effective training datasets Research and implement innovative data curation methods Collaborate with researchers and engineers on data pipelines Requirements:
Strong software engineering skills with Python proficiency Experience building data pipelines and handling large-scale datasets Familiarity with curriculum learning, data mixing, and attribution Experience with Spark, Beam, Pandas, or similar frameworks Knowledge of data quality assessment and mixture experimentation Use Python, data curation, and pipeline engineering to strengthen Cohere’s language model training systems.