ML Engineer, Trust & Safety [Hapiko]

Spin Master is a global children's toy, entertainment and digital gaming company, with Hapiko developing a voice-activated sticker printer for children. The ML Engineer, Trust & Safety will own and improve the generative AI content moderation pipeline, evaluation systems and tooling, ensuring safe product experiences while minimizing over-blocking. The role will also support red teaming, incident response, compliance, and collaboration with ML and backend/platform engineering teams.

Responsibilities:

• Own and maintain content moderation end to end, from what a child says to the image that prints, including copyrighted and licensed characters
• Maintain and grow our safety eval suites, and use them to decide when a model, prompt or threshold change is ready to ship
• Continuously improve our classifiers, and decide where in-house models and third-party models fit best
• Tune moderation over time to cut over-blocking without weakening protection
• Run ongoing red teaming against our models and pipeline
• Maintain review tooling for flagged content and parent reports
• Own the incident response runbook for safety issues
• Maintain the technical side of our compliance obligations, including data retention and deletion
• Partner with our ML and Backend / Platform engineers so safety is built into new models and features before they ship

Requirements:

• Applied safety engineering experience on a product with real users
• You've built evals or classifiers for harmful content, not just configured a vendor API
• Strong software engineering skills in Python: you can ship production code in the moderation path, where it has to be both fast and fail-safe
• Comfort with precision/recall tradeoffs, and the judgment to set thresholds when both kinds of error have a cost
• Good judgment under pressure