Trust & Safety Analyst; III

Responsible AI Scaled Testing Lead

The Responsible AI Scaled Testing Team within Trust & Safety performs pre-launch structured testing for Google's AI applications against safety, fairness and neutrality policies and standards. It is a global team with Responsible AI domain expertise and diverse backgrounds in operations, strategy, ethics, risk management, product management, and program management.

Lead structured pre-launch safety, neutrality and fairness testing, end to end, for Google's GenAI products. For each launch, this will involve defining the standards applicable to the product, defining and executing prompt generation strategies, collaborating with product teams to scrape responses, working with our extended workforce to execute prompt/response rating against defined standard (incl. providing clear instructions, clarifying gray area cases, and/or providing quality calibrations), and conducting in-depth quantitative and qualitative analysis of results.

Intake and triage new launch submissions; understand requirements and kick off engagement or schedule for future start date. Move existing launches through the pre-launch testing process, using Buganizer bugs to track progress. This may include conducting work independently or collaborating with stakeholders to gather information or ensure they are taking action on their end. Aligning on the safety, neutrality and fairness standards applicable to the product, with an eye for driving consistency across product areas.

Translating standards into clear guidelines that can be used to evaluate whether the product's output is compliant with standards. Defining and executing prompt generation strategies to develop a set of prompts that will sufficiently test product compliance with standards. This may entail leveraging LLM-based prompt generation tools and/or defining and providing clear instructions to vendor teams. Collaborating with product teams to scrape responses.

This may entail providing consultation for how to develop a scaled scraping solution (UI, API, etc.), getting access to the model/UI and performing scrapes, and/or defining and providing clear instructions to vendor teams. Executing prompt/response rating against defined standards. This may entail providing clear instructions to vendor teams, clarifying gray area cases, and/or providing quality calibrations. Deep dive analysis:
Conduct in-depth quantitative and qualitative analysis of results, including unexpected, interesting and edge cases, providing clear and actionable insights to inform decision-making around pre- and post-launch mitigation steps.

Bachelor's degree or equivalent practical experience. 4 years of experience in any one of data analytics, Trust & Safety, policy, cybersecurity, or related fields. Experience using data to provide solutions and recommendations. Excellent communication and presentation skills (written and verbal) and the ability to influence cross-functionally at various levels.