Post-Training Applied Researcher

About this role

Post-train open-source language models to outperform frontier closed models on specialized customer tasks. The role combines applied research with production delivery, translating domain-specific data and requirements into training pipelines, reward functions, environments, and evaluations.

About the company

Baseten is an inference cloud that helps companies run and scale AI models in production with flexible infrastructure and developer tooling.

The team

EPD, Training, Post-Training

Stack

What you'll do

• Design and run post-training pipelines, including supervised fine-tuning, reinforcement learning, reward engineering, and synthetic data generation.
• Build task-specific training environments and evaluations for multi-turn tool use and agentic workflows.
• Work with customers to turn production data and usage patterns into training signals and reward loops.
• Analyze experiments and diagnose reward hacking, importance-sampling drift, and advantage-estimation instability.
• Publish research findings and contribute to open-source training libraries.

What we're looking for

• Hands-on LLM reinforcement-learning experience and a deep understanding of GRPO or PPO objectives and design.
• Strong reward-engineering judgment, including anticipating reward exploitation at scale.
• Experience building multi-turn agent environments with tool use.
• Comfort across dataset construction, training, evaluation, and deployment.
• Production ML experience; preference for having closed a production-data-to-model-improvement loop.

Nice to have

• Experience with RL training frameworks.
• Publications at NeurIPS, ICML, or ICLR focused on LLM reinforcement learning, reward modeling, or alignment.

Benefits

• Meaningful equity.
• 100% medical, dental, and vision coverage for U.S. employees and dependents.
• Flexible PTO and a company-wide winter break.
• Paid parental leave.
• Fertility and family-building stipend.
• Company-facilitated 401(k) for U.S. employees.