Principal Research Engineer, Model Training  Post-Training

📋 Description

• Own the model-improvement roadmap across capability, reliability, EI, cost.
• Lead training and post-training strategy (SFT, RLHF, RLAIF, reward modeling).
• Drive architecture/optimization for transformer-based and hybrid models.
• Lead large-scale training on distributed GPU clusters (1000+ GPUs).
• Define data strategy: curation, deduplication, decontamination, human-in-the-loop.
• Build evals and release-quality systems (evals, gates, readiness).

🎯 Requirements

• Experience leading large-scale LLM or foundation-model training/post-training.
• Deep experience with transformer models, hybrid architectures, distributed training.
• Post-training methods: SFT, RLHF, DPO, GRPO, RLAIF, reward modeling.
• Operate large GPU clusters (1000+ GPUs) and distributed training infra.
• Strong systems instincts around throughput, cost, reliability, observability.
• Judgment on data quality, evaluation design, release readiness, and model behavior.
• PhD in CS/ML/AI or equivalent practical experience.

🎁 Benefits

• Diverse medical, dental and vision options
• 401k matching
• Unlimited paid time off
• Parental leave and caregiver flexibility
• Visa support for international Bay Area employees