Project Manager, Abusive Actor Operations

Meta Meta · Big Tech · Austin, TX

This role focuses on AI governance and oversight for integrity and support use cases. The Project Manager will act as a human expert, making decisions on complex cases, defining how AI/ML models are reviewed and audited, and setting human-in-the-loop thresholds. They will also establish and steward expert-labelled datasets for training and evaluating models, ensuring data quality and model accuracy. The role involves analyzing model and operational data to identify trends and improve enforcement, and partnering with engineering and data science to enhance labeling interfaces and oversight tooling. A key aspect is defining governance frameworks, guardrails, and compliance with policy, regulatory, and legal requirements.

What you'd actually do

  1. Serve as a human expert and final decision-maker on ambiguous, high-severity, and precedent-setting cases, and adjudicate escalations involving potentially abusive content or real-world threats
  2. Own AI oversight and governance for integrity and support decision-making: define how AI and ML models are reviewed, audited, and held accountable, and set the human-in-the-loop and human-out-of-the-loop thresholds that govern when models can act autonomously
  3. Establish and steward expert-labelled ("golden") datasets used to train, evaluate, and calibrate AI/ML models, owning label quality, precision/recall, model-agreement measurement, and drift over time
  4. Define and enforce governance frameworks (quality standards, guardrails, and escalation paths) that ensure model decisions and data labels comply with all policy, regulatory, and legal requirements
  5. Analyze and audit model and operational data to identify user trends, malicious actor behavior, systematic model errors, and coverage gaps, and translate findings into recommendations that improve enforcement accuracy at scale

Skills

Required

  • Experience in owning quality, decision-making, or oversight processes, setting standards and exercising authoritative judgment on complex cases
  • Experience communicating with a variety of audiences, including engineering, data science, and technical teams
  • Demonstrable experience using sound judgment and exhibiting Demonstrable experience using sound judgment and maintaining composure when prioritizing competing demands based on impact and urgency
  • Familiarity with AI governance, trust and safety, or content/integrity policy
  • Knowledge of programming languages such as Python, R, or SQL
  • Experience with AI/ML oversight, human-in-the-loop workflows, model evaluation, or annotation/labeling quality
  • Strategic thinker with strong analytical and problem-solving skills
  • 2+ years of experience working with data tools such as SQL, SAS, and data visualization (e.g., Tableau) tools

Nice to have

  • Familiarity with Meta's family of products and their user experience

What the JD emphasized

  • AI oversight and governance
  • human-in-the-loop
  • expert-labelled datasets
  • model evaluation
  • policy, regulatory, and legal requirements
  • guardrails

Other signals

  • AI oversight and governance
  • human-in-the-loop
  • expert-labelled datasets
  • model evaluation
  • policy, regulatory, and legal requirements