Research Scientist, RL Training

Job Description

Research Scientist - Reinforcement Learning for LLMs

Location: Redwood City, CA

Work Model: On-site, Remote

About The Role

Snorkel is seeking a Research Scientist to focus on reinforcement learning for training and aligning large language models (LLMs). This foundational research role aims to solve the open data problem of generating reliable training data, reward signals, and procedures to steer LLM behavior.

Main Responsibilities

• Research and implement reinforcement learning techniques (GRPO, RLHF, RLAIF, DPO, reward modeling) to create data products for LLM training and fine-tuning.
• Design and build data pipelines for generating high-quality training signals and improving model generalization.
• Prototype and iterate on end-to-end RL training recipes.
• Collaborate with research, engineering, and delivery teams to translate RL research into customer-ready data products.
• Stay current with advancements in LLM training, alignment research, and scalable RL methods.
• Contribute to research publications and internal knowledge base.
  
  

Preferred Qualifications

• Deep expertise in reinforcement learning from human or AI feedback, reward modeling, and credit attribution.
• Experience training or fine-tuning 30B LLMs at scale, including distributed training infrastructure.
• Strong proficiency in Python and ML frameworks (PyTorch, HuggingFace) and RL frameworks (Verl, SkyRL).
• Solid software engineering fundamentals for building reproducible research prototypes.
• Familiarity with ML infrastructure and cloud platforms (AWS, GCP, Kubernetes, Slurm); experience with large-scale RL training pipelines is a plus.
• Comfort with high-iteration, open-ended research environments and shifting constraints.
• Ph.D. in machine learning, reinforcement learning, or related field, or equivalent industry experience.
  
  

Salary Range

$200,000 - $275,000 USD