Research Engineer

Metis, Inc.; San Francisco (CA); On-site; USD 200,000 - 1,000,000; Full time

Benefits offered by this job

Full medical, dental, and vision insurance

Wellness & Learning stipend

Unlimited Doordash meals

$25,000 housing stipend

Job summary

Metis, Inc. is seeking a Research Engineer to develop the next generation of autonomous post-training systems using their Mantis platform. The role involves designing and executing large-scale experiments, contributing to research, and collaborating with engineering teams to improve AI learning.

Candidates must have deep experience in machine learning, particularly in reinforcement learning, and a strong proficiency in Python along with ML frameworks. The position offers a base salary range of $200,000 to $1,000,000, full medical benefits, wellness stipends, and daily meals.

Qualifications

• Deep experience in machine learning, preferably in reinforcement learning or alignment.
• Demonstrated research contributions and ideally published papers.
• Strong proficiency in Python and ML frameworks.

Responsibilities

• Research and build an autonomous post-training agent leveraging the Mantis platform.
• Design and execute large-scale experiments on synthetic data generation.
• Collaborate cross-functionally to deploy and evaluate models.

Skills

Machine learning

Reinforcement learning

Python

ML frameworks (PyTorch, JAX, or TensorFlow)

Distributed training

Job description

As a Research Engineer at Metis, you’ll work on building the next generation of autonomous post-training systems that leverage our Mantis platform. You’ll operate at the intersection of cutting-edge ML research and scalable engineering, designing, implementing, and deploying algorithms that improve how AI agents learn from feedback, synthetic data, and real-world interactions.

You’ll move seamlessly between papers and production, leading large-scale experiments, creating optimized training pipelines, and helping shape the future of post-training autonomy. You’ll have significant ownership, high compute budgets, and the mandate to push the state of the art in applied reinforcement and preference optimization.

What You’ll Do

• Research and help build an autonomous post-training agent leveraging the Mantis platform
• Design and execute large-scale experiments on synthetic data generation and algorithmic architecture
• Develop and refine methods for reinforcement learning, reward modeling, and human feedback integration
• Collaborate cross-functionally with Core and Platform Engineering to deploy and evaluate models in production settings
• Publish or contribute to leading-edge research in the post-training domain
• Use tooling and compute efficiently to iterate on experimental pipelines and accelerate research velocity

Requirements

• Deep experience in machine learning, preferably reinforcement learning, post-training, or alignment research
• Demonstrated research contributions; ideally published papers (ICML, NeurIPS) or public implementations
• Strong proficiency in Python and ML frameworks (PyTorch, JAX, or TensorFlow)
• Comfort with distributed training, high-throughput data pipelines, and large-scale experiment management
• Ability to reason independently, formulate hypotheses, and run experiments from idea to insight to product impact
• Base: $200,000-$1,000,000
• Full medical, dental, and vision
• Wellness & L&D stipend
• Breakfast, lunch, and dinner provided (Unlimited Doordash)
• $25,000 housing stipend