SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
MagicSchool is a generative AI platform for teachers with over 7 million users globally. You will lead the Trust & Safety function as a Director-level role, owning all aspects of AI safety, responsible deployment, and incident response for a product serving educators and students worldwide.
Key Responsibilities:
- Own the safety of AI output to teachers and students, including policy development, model-level guardrails, evaluations, red-teaming, and incident response.
- Establish and maintain trust posture with school districts through safety documentation, RFP responses, and alignment with state AI-in-education guidance.
- Integrate safety evaluations and monitoring into product infrastructure, ensuring each new tool includes safety acceptance criteria during scoping rather than at launch.
- Manage student-facing safety specifically: age-appropriate content, self-harm and crisis handling, and escalation paths to educators.
- Run the safety incident process end-to-end and own the postmortem feedback loop into product and engineering teams.
- Partner with Legal on COPPA, FERPA, state privacy laws, and emerging AI regulation, with T&S owning operational reality and Legal owning legal position.
- Build and lead a small team (1–3 initially), scaling through 2027.
You will work with a multi-provider LLM stack (OpenAI, Anthropic, Google via GCP Vertex), custom classifiers, Datadog for observability, and incident management tools (Linear, Slack, Notion). The product is built on TypeScript/Next.js on Vercel with Supabase PostgreSQL backend.
REQUIREMENTS:
Must Haves:
- Proven ownership of trust & safety or responsible AI for a consumer or youth-facing product at scale, with real incidents, real policy decisions, and regulator interactions.
- Hands-on expertise with LLM safety: evals, classifiers, prompt/system-level guardrails, red-teaming. Ability to read eval results and determine what to change.
- Track record of making safety a velocity enabler, not friction—building guardrails as infrastructure so teams ship faster with confidence.
- Ability to make judgment calls under ambiguity and own them publicly to districts and executive team.
- Strong writing skills: safety documentation, district-facing explainers, incident communications.
- Hypothesis-first mindset; escalates one-way doors quickly rather than delaying decisions.
Nice to Haves:
- K-12 or youth-platform experience (GoGuardian, Securly, Roblox, Discord, YouTube Kids, Epic Games).
- Familiarity with state AI-in-education frameworks and district procurement questionnaires.
- Prior work with child-safety NGOs or standards bodies (Common Sense Media, NCMEC, Thorn, ISTE).
- Data/ML fluency to partner directly with engineering on classifiers and eval pipelines.
- Comfort being the external face of safety at conferences, with press, and in front of district boards.
