AI Safety Operations

View full description

Requirements:

• Background in safety operations, content moderation, customer support, and/or product management
• Experience designing, and implementing operational processes & protocols for human or automated implementation. Experience doing this in trust and safety, AI data labeling, customer support or security teams strongly preferred.
• Strong track record working with data. You have experience crafting SQL queries and tables and working with cloud data warehouses (e.g. Snowflake, Big Query). You can manipulate, understand and draw insights from structured or unstructured data.
• Track record managing operational teams.
• Operational rigor: You thrive under pressure, staying organized and decisive even when handling emergencies or sudden surges in risky content. You build foundations that can handle high volumes with a bias towards automations/scalability.
• Cross-functional fluidity: You’re skilled at collaborating with engineers, product teams, and other stakeholders to ensure safety measures integrate seamlessly with ongoing innovations. You do this by pointing towards a common goal, even when the safety requirements might be perceived to slow down.
• Thick skin: You maintain composure and objectivity while confronting disturbing or complex material, ensuring it doesn’t impede your (or your teams’) judgment or well-being.

Responsibilities:

• Design and implement protocols to handle various safety operational workflows including moderation, appeals, and safety-related user requests.
• Mentor, and lead a group of safety ops specialists who review & assess a high volume of flagged content, handle appeals, investigate and respond to incidents, and support our machine learning engineering team.
• Manage customer requests and safety incidents.
• Oversee our response to internal and external safety requests. Ensure quality and consistency of our response.
• Support development of automated safety systems.
• Work with engineering, policy and investigations team members to support building automated systems, including data labeling for moderation classifiers & detection models.
• Craft clear guidelines: Translate our prohibited use policies into operational guidelines for human reviewers and automated systems.
• Build processes with an eye to automation; implement simple automations independently
• Stay abreast of developments in trust and safety ,and AI safety, and introduce ideas for automation and efficiency to the team. Survey the ecosystem for tooling, processes and standards that can improve safety outcomes, and support selection, integration and management of external tools.