Research Scientist / Research Engineer, Safety + Alignment, Search

About the Role

We’re looking for a Researcher to lead our safety and alignment research: building the systems that distinguish useful web content from spam, manipulation, and other adversarial or low-quality material.

The challenge extends beyond evaluating individual pages. A page may appear credible in isolation, while its links, metadata, or relationships with other sites reveal coordinated manipulation. As AI makes convincing content easier to produce at scale, detection systems must adapt to increasingly sophisticated adversaries.

You’ll define the technical approach, build classifiers and evaluations, and shape how integrity signals influence what Search indexes and retrieves. Working closely with Search and broader Safety and Integrity teams, you’ll use frontier models to advance detection methods and bring them into production.

Responsibilities

• Define the technical direction for Safety + Integrity, identifying gaps and determining which problems and interventions to prioritize.
• Develop AI and ML systems to detect spam, scams, phishing, malicious domains, prompt injections, and other adversarial or low-quality web content.
• Discover useful signals across page content, links, domains, metadata, and relationships between websites, including patterns that reveal coordinated manipulation.
• Build evaluations that measure detection quality and how integrity interventions affect the information Search surfaces.
• Translate detection signals into indexing, ranking, and filtering decisions, improving what models can retrieve and rely on.
• Serve as the technical owner for integrity within Search, partnering with researchers, engineers, and broader Safety and Integrity teams to build and deploy these systems.

You Might Thrive in This Role If You Have

• Deep experience in web spam, search quality, adversarial content detection, or related integrity problems at scale.
• Strong applied ML and research skills, with experience developing classifiers, detection methods, and evaluations for real-world systems.
• An understanding of how adversaries manipulate content and links, coordinate across sites, and adapt to detection systems.
• A track record of independently defining ambiguous technical problems and driving solutions from experimentation through production.
• Experience working across research, engineering, and integrity teams while remaining hands-on with technical work.
• A first-principles approach to using frontier models and compute, and a strong interest in keeping detection systems effective as adversaries evolve.