Product Analyst, AI Safety & Guardrails

Dynamo AI; San Francisco (CA); On-site; USD 110,000 - 150,000; Full time

Job summary

Dynamo AI in San Francisco seeks an analytical PM/analyst to own design and specification of items in multiple streams: scope the problem, design experiments, define success metrics, run analyses, and deliver recommendations. You may rapidly build small AI-assisted tools to test ideas and unblock the team.

You won’t write training or infrastructure code; your focus is experimental design, data quality, and rigorous evaluation as a path toward a senior, technical product role.

Qualifications

• Requires a degree with strong quantitative or analytical content.
• Demonstrated framework for thinking through problems.
• Good instinct for data, ML concepts, and trustworthy experiments.
• Proficiency in Python and pandas for working with data.
• Interest in AI security or safety and adversarial thinking.
• Attention to detail and clear writing.

Responsibilities

• Guardrail training data: curate and design data for training an SLM guardrail and ensure labeling consistency as datasets grow.
• Orchestration experiments: test different guardrails and control configurations to identify the best performing setup.
• Synthetic data quality: assess fidelity and diversity of synthetic training/eval data.
• New evaluations: research and scope new evaluation methods for guardrails and AI applications, then specify how they work.
• Rapid tooling: build small custom AI tools to test hypotheses or unblock the team.
• Benchmarking: measure performance using standard metrics (FNR, FPR, precision, recall) and report findings.

Skills

Python

Pandas

Data analysis

Experimental design

Analytical thinking

Education

Degree with quantitative/analytical content

Job description

Dynamo AI in San Francisco seeks an analytical PM/analyst to own design and specification of items in multiple streams: scope the problem, design experiments, define success metrics, run analyses, and deliver recommendations. You may rapidly build small AI-assisted tools to test ideas and unblock the team.

You won’t write training or infrastructure code; your focus is experimental design, data quality, and rigorous evaluation as a path toward a senior, technical product role.