The company is a small AI team that builds post-training methods and infrastructure for AI agents running in production.
You will work across research and engineering to build systems that raise model quality and cut cost and latency in real applications.
Responsibilities
Create evaluation suites, graders, and data pipelines based on production workflows and traces.
Take ownership of post-training work for customers, covering data preparation, evaluations, reinforcement learning environments, and training runs.
Build agent environments and harnesses that make experiments reproducible and support training data generation.
Run and analyze post-training experiments, reproduce promising research, and debug the systems involved.
Convert experiments that succeed into dependable parts of a post-training and continual learning platform.
Requirements
Between 1 and 5 years of engineering experience at an AI startup or in an ML research group, working directly with large language models.
Hands-on experience training, evaluating, and serving language models, not only calling model APIs or building prompt and agent wrappers.
Strong Python skills and a solid grounding in machine learning and computer science.
Hands-on experience with tools such as PyTorch, LLM training, evaluation harnesses, or model serving.
A substantial project, research contribution, shipped system, internship, or open-source work that shows technical depth and ownership.
Interest in or experience with post-training or inference, and the ability to learn unfamiliar systems, debug on your own, and explain technical decisions clearly.
A habit of measuring carefully and following through, with a focus on confirming that systems work in practice.
Conditions
Base salary ranges from $200,000 to $300,000 per year.
Visa sponsorship is available.
The role is on-site in Mountain View, California, United States.