Production-Ready LLM RL Researcher (Post-Training)

US Health Partners, LLC; San Francisco (CA); Vor Ort; USD 180.000 - 240.000; Vollzeit

Benefits dieser Stelle

Competitive compensation

Equity

Medical, dental, and vision insurance

Flexible PTO including Winter Break

Paid parental leave

401(k)

Learning and networking opportunities

Zusammenfassung

Baseten is seeking researchers focused on post-training AI advances to work with leading AI companies. You will fine-tune LLMs, design reward loops, and build training pipelines that translate real customer data into production-ready models.

You will iterate on reward engineering, multi-turn tool-using environments, and scalable evaluation frameworks, shipping models to millions of users.

Qualifikationen

• Hands-on experience training LLMs with reinforcement learning (e.g., PPO/GRPO) and producing robust results.
• Ability to translate customer domain requirements into a training curriculum for models.
• Experience building multi-turn agent environments with tool use beyond single-turn QA.
• Comfort across the full pipeline from dataset construction to training, evaluation, and deployment.
• Experience with production ML systems and closed a training–inference loop where production data informs improvement.

Aufgaben

• Design and run post-training pipelines (SFT, PPO, RLHF) and reward function engineering.
• Build task-specific training environments and evals for domains like healthcare and code generation.
• Collaborate with customers to convert production data into training signals and manage distribution shift.
• Run and analyze experiments end-to-end, diagnosing reward hacking and estimation instabilities.
• Publish findings and contribute to Baseten's open-source training libraries.

Kenntnisse

LLM fine-tuning

Reinforcement learning

Reward engineering

Multi-turn environments

Production ML systems

Tools

RL training frameworks

Jobbeschreibung

Baseten is seeking researchers focused on post-training AI advances to work with leading AI companies. You will fine-tune LLMs, design reward loops, and build training pipelines that translate real customer data into production-ready models.

You will iterate on reward engineering, multi-turn tool-using environments, and scalable evaluation frameworks, shipping models to millions of users.