- Spotify
- New York, NY
- Full-Time
- 3 days ago
Staff Machine Learning Engineer – Policy & Safety: our view in 3 lines...
- The Role: Build and scale machine learning systems for proactive content detection, policy evaluation, and safety across multimodal signals for policy and safety use cases.
- The Person: Design, implement, and architect scalable ML systems and policy evaluation frameworks for content detection and safety, drive feedback loops for continuous model improvement, translate regulatory requirements into ML designs, and mentor other engineers.
- Requirements: Experience building and shipping production-grade machine learning systems at scale, expertise in ML evaluation including dataset design, metrics, model performance monitoring, and work with multimodal models and human-in-the-loop systems.
Job Description
We design Spotify’s consumer experience—end to end, moment to moment, across every screen, platform, and partner integration. Our mission is to make listening feel effortless, personal, and joyful for billions of users around the world. That means turning complexity into clarity across hundreds of touchpoints—from our mobile and desktop apps to the smart speakers, TVs, cars, and integrations where Spotify shows up every day. If it touches a consumer, we shape it. We bring deep insight into human behavior, design, and technology to craft experiences that feel intuitive, expressive, and unmistakably Spotify.
About the Team
The Policy & Safety team sits within the Content Platform domain and builds the systems that keep Spotify safe and trustworthy at scale. We own the infrastructure behind content moderation, including detection models, policy enforcement systems, compliance pipelines, and the safety-by-default platform.
Our work sits on the critical path of every new content type and product experience—from messaging and comments to collaborative and agentic features. We partner closely with Trust & Safety, Legal, and Public Affairs to ensure that as Spotify evolves, safety is built in from the start—not added later.
