Lead, Generative Media Evaluation & Quality

Vamp; Los Angeles (CA); On-site; USD 193,000 - 269,000; Full time

Job summary

Meta seeks a Content Engineer to own the human evaluation layer for image and video generation, defining rubrics, running eval rounds, and translating subjective judgments into measurable signals.

The role partners with product, research, and modeling teams to set quality standards, determine when to automate, and drive model behavior decisions. A strong background in video production and evaluation frameworks is required.

Qualifications

• 10+ years of experience in video production, content strategy, editorial standards, media evaluation or relevant fields.
• Experience authoring evaluation rubrics for video and/or image quality, including dimension design, rating scales, tie-breaking rules and annotator guidelines
• Hands-on video production expertise, direction, shooting, editing, post and/or VFX, with the craft vocabulary to diagnose quality failures precisely
• Experience evaluating generative media output (text-to-video, image-to-video, video-to-video, text-to-image) and diagnosing model failure modes
• Experience partnering with AI/ML research or modeling teams, translating evaluation results into training, data or model-behavior recommendations
• Experience running human evaluation programs at scale: rater calibration, inter-rater reliability, golden sets, quality audits
• Experience leveraging quantitative insights to inform content and quality decisions
• Proven ability to build influence and drive alignment across product, engineering, design, research and analytics partners
• Experience leading and/or mentoring others
• Bachelor's degree in a related field, or equivalent practical experience

Responsibilities

• Own the evaluation rubrics for image and video generation models, define the quality dimensions, scales and decision rules, and maintain them as models and capabilities evolve.
• Design and run human eval rounds that produce training and fine-tuning signal, including prompt set construction, golden sets and side-by-side model comparisons.
• Set and defend the launch quality bar for generative media features, and make clear go / no-go quality recommendations.
• Calibrate and manage rater pools and vendor teams
• monitor inter-rater reliability and drive measurable improvements in agreement.
• Partner with research and modeling teams to translate eval findings into model, data and post-training priorities.
• Scale human judgment into automated measurement (LLM-as-a-judge, auto-metrics) and validate automated scores against human ground truth.
• Lead through collaboration, mentoring other content engineers on eval craft and rubric design.
• Stay ahead of the generative media landscape, benchmarking against competitive and open-source model output.
• Deliver high-quality evaluation work on model and launch timelines in a fast-paced, dynamic environment.

Skills

Ten+ years experience

Rubrics design

Video production

Generative media evaluation

Partner with ML/AI teams

Large-scale eval programs

Quantitative insight

Mentoring others

Education

Bachelor's degree in a related field

Equivalent practical experience

Tools

Annotation platforms

Eval tooling

Data pipelines

Job description

Meta seeks a Content Engineer to own the human evaluation layer for image and video generation, defining rubrics, running eval rounds, and translating subjective judgments into measurable signals.

The role partners with product, research, and modeling teams to set quality standards, determine when to automate, and drive model behavior decisions. A strong background in video production and evaluation frameworks is required.