AI Research Engineer

Cerebro; San Francisco (CA); On-site; USD 270,000 - 330,000; Full time

Benefits offered by this job

Equity

Job summary

Frontier AI in San Francisco is seeking a Member of Technical Staff – Research to advance frontier-model evaluations, environments and reward mechanisms. You will work on designing challenging benchmarks, building interactive environments and scalable rewards to drive model improvement.

The role focuses on research and infrastructure development, not mere dashboarding, with equity upside and a highly technical team led by former Meta AI researchers.

Qualifications

• LLM evaluation, benchmark development or model capability assessment.
• Agentic systems, tool-use environments or multi-step reasoning.
• Reinforcement learning, RLHF, GRPO, reward modelling or verifiable rewards.
• Synthetic data generation, automated evaluation or executable verification.
• Post-training research and evaluation-to-training pipelines.

Responsibilities

• Design agent evaluations and benchmarks exposing frontier models’ limits.
• Develop interactive RL environments with meaningful learning signals.
• Explore scalable reward design and verification techniques.
• Use evaluation results to inform data generation, training and model improvement.
• Investigate frontier-model capabilities, generalisation and robustness.

Skills

LLM evaluation

Benchmark development

Reinforcement learning

Research engineering

Job description

Member of Technical Staff – Research | Frontier AI

San Francisco | Up to $300,000 Base + Equity | 5 Days Onsite

What happens when the benchmarks we use to measure AI are no longer sufficient to improve it?

We're partnering with a well-funded San Francisco AI research company tackling one of the most important challenges in frontier AI: building the evaluations, environments and feedback mechanisms that enable increasingly capable models to learn, reason and improve.

As frontier models become more sophisticated, static benchmarks are becoming less informative. The next generation of AI development requires richer environments, more reliable reward signals and evaluations that capture how agents actually behave across complex, multi-step tasks.

That's exactly what this team is building.

They're looking for a Member of Technical Staff – Research to help push this work forward.

The research

You'll work on problems spanning:

• Agent evaluations and benchmarks: Designing challenging evaluations that expose the limitations of frontier models across reasoning, tool use and complex agentic workflows.
• RL environments: Developing interactive, verifiable environments that generate meaningful learning signals for increasingly capable agents.
• Reward design: Exploring scalable reward mechanisms, automated verification and techniques for assessing complex model behaviour.
• Evaluation-driven post-training: Using model failures, trajectories and evaluation results to inform data generation, training and model improvement.
• Frontier-model research: Investigating emerging capabilities, generalisation, robustness and the limitations of existing evaluation methodologies.

This isn't a role focused on maintaining dashboards or simply measuring model performance.

It's an opportunity to help build the infrastructure and research foundations that determine how the next generation of AI systems are trained and evaluated.

Who we're looking for

We're particularly interested in researchers and research engineers who have worked on one or more of the following:

• LLM evaluation, benchmark development or model capability assessment.
• Agentic systems, tool-use environments or multi-step reasoning.
• Reinforcement learning, RLHF, GRPO, reward modelling or verifiable rewards.
• Synthetic data generation, automated evaluation or executable verification.
• Post-training research and evaluation-to-training pipelines.Designing environments for coding agents, computer-use agents or other complex AI systems.

You might currently be working in a frontier AI lab, an AI evaluation company, a post-training research team or an academic research group.

We're much more interested in the originality and technical depth of your work than your current job title.

Why this opportunity?

Work directly on research problems at the intersection of frontier-model capabilities, evaluation and reinforcement learning.

Join a highly technical team founded by former Meta AI researchers.

Help shape research direction rather than simply execute an established roadmap.

See your research translated into systems used to develop and improve advanced AI models.

Competitive compensation of up to $300,000 base salary plus equity.

If you've been building benchmarks, studying agent failures, designing RL environments or thinking deeply about how evaluation can drive model improvement, I'd be very interested in speaking with you.