Open-Model Safety & Eval Program Manager

Reflection AI Ltd; New York (NY); On-site; USD 140,000 - 200,000; Full time

Benefits offered by this job

Top-tier comp

Stock options

Health plan

Meals provided

Parental leave

Unlimited PTO

Visa sponsorship

Team events

Job summary

Reflection AI Ltd. is seeking a Research Program Manager to embed with research and infrastructure teams, accelerating frontier model development.

This foundational role will define evaluation frameworks, build safety operations, and connect evals to the model lifecycle in a fast-moving, high-stakes environment. You will drive scoping of eval infrastructure, engage with researchers and engineers, and establish processes that scale as reflection interfaces with the broader safety ecosystem.

Qualifications

• 7+ years in technical program management, research operations, or ML engineering.
• Experience standing up new functions or programs from scratch.
• Familiar with model evaluation and AI safety landscapes.
• Able to engage with researchers and engineers on model behavior and data pipelines.

Responsibilities

• Build the foundational infrastructure for model evals and safety.
• Stand up model safety operations with workflows and cadences.
• Embed safety checkpoints into development processes across training stages.
• Establish engagement with external safety ecosystem and partnerships.
• Provide visibility through reporting on model safety status and risks.

Skills

Technical program management

Research operations

ML engineering

Job description

Reflection AI Ltd. is seeking a Research Program Manager to embed with research and infrastructure teams, accelerating frontier model development.

This foundational role will define evaluation frameworks, build safety operations, and connect evals to the model lifecycle in a fast-moving, high-stakes environment. You will drive scoping of eval infrastructure, engage with researchers and engineers, and establish processes that scale as reflection interfaces with the broader safety ecosystem.