Post-Training Research Scientist

present of AI; San Francisco (CA); On-site; USD 160,000 - 260,000; Full time

Job summary

Baseten is seeking a researcher who can identify meaningful open problems in post-training and inference efficiency, design rigorous experiments at scale, and hand findings to engineers who ship them. The ideal candidate has deep ML research experience with strong publications and can translate ideas from theory to working code with measurable results, impacting production systems.

You will work on distillation, memory handling for long-running agents, and integrate results on a platform serving

Qualifications

• Deep machine learning research experience with strong publications.
• First-author ML publications.
• Post-training and RL experiments.
• Open-source research contributions.
• Research that ties to production impact.

Responsibilities

• Identify worthwhile post-training and inference efficiency problems.
• Design and run careful large-scale experiments.
• Translate findings into actionable improvements for engineers.
• Reproduce recent work on distillation or cache compression with open models.
• Write clear, reproducible results and documentation.

Skills

ML Research

Publications

Experiment Design

Production impact

Job description

The role is for a researcher who can pick worthwhile open problems in post-training and inference efficiency, run careful experiments at large scale, and then hand findings to engineers who ship them. A strong candidate has deep machine learning research experience with strong publications, and can also carry an idea from theory to working code and measured results. Good judgment about which questions change how systems get built, not just which move a benchmark, matters a lot here.

How it fits the AI era

The work is itself about how models learn, align and run efficiently, so AI is the subject as much as the tool. Day to day, AI tools can speed up coding, experiment setup and literature review. Choosing problems, designing clean experiments, and judging whether a result really holds up remain human work.

How it builds AI experience

Someone in this role would work directly on post-training methods, distillation, and memory handling for long-running agents on a platform that serves fast-growing AI companies. That means hands-on time with very large models across many machines and real production traffic. To build toward it, reproduce recent work on distillation or cache compression using open models and write up clear results.

Worth showing

• First-author ML publications
• Post-training and RL experiments
• Open-source research contributions
• Research tied to production impact

This is our own read, written from the employer's posting and the company's Spotlight. The employer's posting is the source of truth for the role.

About Baseten

High-performance cloud inference platform for deploying, serving, and scaling open-source and custom artificial intelligence models.