Research Engineer, Mid-Training

About this role

Join an applied AI lab as a Research Engineer focused on mid-training—the critical phase between pre-training and post-training where base model capabilities are refined. You'll own data strategies, training schedules, synthetic data pipelines, and capability improvements that directly shape what AI software agents can accomplish.

What you'll do

• Design and iterate on high-quality data mixtures for late-stage training runs, developing methods for sourcing, filtering, and weighting data
• Drive targeted capability improvements in coding, mathematics, and reasoning through curated data strategies and training interventions
• Develop and evaluate synthetic data pipelines that generate training signal at scale, understanding production failure modes
• Research and optimize multi-stage learning rate schedules, warmup strategies, and compute allocation across training phases
• Implement and research methods for extending effective context length without degrading short-context performance
• Build evaluations that distinguish real capability improvements from benchmark overfitting and close the loop on training decisions

What they're looking for

• End-to-end LLM training pipeline knowledge (pre-training, optimization, architecture, mid/post-training interactions)
• Hands-on experience with continual pre-training, annealing, or late-stage data mixing for large models
• Data quality curation and filtering at scale
• Synthetic data pipeline development and evaluation
• Python and PyTorch proficiency with distributed training debugging
• Optimization, statistics, and machine learning theory fundamentals
• Ability to distinguish real effects from noise and overfitting
• Comfort with ambiguous, fast-moving research environments

Benefits

• Access to substantial compute resources with training jobs running across thousands of GPUs
• Small, highly selective team where research and product move together
• Prototypes reach real deployment quickly
• Minimal process overhead in a fast-moving environment
• Opportunity to work on one of AI's most competitive problems