Research Program Manager - Model Evals and Safety

Our Mission

Reflection’s mission is to build open superintelligence and make it accessible to all
.

We’re developing open weight models for individuals, agents, enterprises, and even nation states. Our team of AI researchers and company builders come from Deep Mind, OpenAI, Google Brain, Meta, Character.

AI, Anthropic and beyond.

Research Program Managers at Reflection are high-leverage leaders and operators who embed directly with research and infrastructure teams to accelerate the pace of frontier model development. They are not project trackers. They are force multipliers who bring clarity to ambiguity, drive decisions when the path forward is unclear, and ensure that the work happening across multiple teams connects into a coherent whole.

This is a foundational role. Reflection is building model evals and safety from the ground up, and this RPM will be at the center of that effort. You won't be stepping into an established function with existing processes and tooling. You will be the person who figures out what this function needs to look like, stands it up, and makes it real.

That means defining the evaluation frameworks, building the operational infrastructure for model safety, establishing the processes that connect evals to the model development lifecycle, and laying the groundwork for how Reflection interfaces with the broader safety ecosystem. This is 0-to-1 work in its purest form.

You bring a first-responder mentality. When things go sideways, you don't wait to be asked. You jump in, assess the situation, cut through noise, align the people who need to be aligned, and drive resolution.

• Build the foundational infrastructure for model evals and safety ine the evaluation frameworks, tooling requirements, and operational processes that will underpin how we assess model capabilities, risks, and readiness for release.

• Stand up model safety operations as a function, including establishing the workflows, review cadences, and decision frameworks that connect safety evaluation to the model development and release lifecycle.

• Partner with research and engineering leads across pre-training, mid-training, and post-training to embed safety and evaluation checkpoints into the development process in a way that is rigorous without being a bottleneck.

• Drive the scoping and prioritization of eval science and eval infrastructure investments, working with technical leads to determine what to build in-house, what to adopt, and where to invest research effort.

• Establish Reflection's engagement with the external safety ecosystem, including third‑party assessments, academic partnerships, and industry safety frameworks. Represent the company's safety posture to external stakeholders with credibility and clarity.

• Create visibility and reporting structures that give leadership a clear, honest picture of model safety status, evaluation coverage, and open risks, so they can make informed decisions at the pace the business requires.

• Champion a culture of blameless post‑mortems and continuous learning, turning every safety‑relevant finding into a concrete improvement to our systems and processes.

• 7+ years of experience in technical program management, research operations, or ML engineering, with demonstrated experience standing up new functions, teams, or programs from scratch.

• Familiar with the landscape of model evaluation and AI safety, including evaluation methodologies, red‑teaming, alignment research, and the evolving regulatory and industry safety ecosystem. You don't need to be a safety researcher, but you need to understand the space well enough to make sound judgments about what matters and what to prioritize.

• Deep enough technically to engage with researchers and engineers on topics like model behavior, evaluation design, data pipelines, and safety‑critical system architecture. You follow the technical thread and you know when something doesn't add up.

• Proven ability to build structures where none exists. You've taken ambiguous mandates and turned them into functioning programs with clear ownership, measurable outcomes, and durable processes.

• Strong stakeholder management skills…