Creator-Safety Operating Model Architect

The mandate

The platform’s move into subscriptions, live tipping and branded collaborations has created safety cases that cross boundaries built for content moderation alone. A report of grooming may involve private messages, payment pressure, an external agency and a creator-support escalation, yet each team applies a different severity model and clock. Executives cannot tell whether slow resolution reflects detection, jurisdiction, evidence preservation or handoff failure. The defined problem is to design one operating system for high-harm creator cases without collapsing policy judgment into queue management.

The principal deliverable is a creator-safety operating-model blueprint covering intake, risk classification, emergency escalation, evidence custody, specialist investigation, payment restriction, victim-centred support, law-enforcement referral, account action, appeal and learning. It will include a case taxonomy, decision-rights matrix, regional service design, protected-data pathways, staffing model, vendor boundaries, quality controls and measures that separate speed from procedural integrity. The blueprint must show how minors, coerced creators and credible imminent harm receive differentiated treatment.

Four milestones govern the five-month engagement. By week four, the first milestone maps fifteen closed cases and reconciles every delay and override. Week nine delivers alternative operating models, demand forecasts and a regional legal-dependency register. Week fifteen produces a controlled tabletop test using grooming, doxxing, impersonation, payment coercion and malicious-report scenarios. At week twenty, the final milestone provides the approved design, implementation backlog, investment case, role charters and a board paper identifying unresolved policy choices.

Acceptance rests with the chief operating officer and Safety Committee, with legal, child-safety and privacy leaders confirming professional boundaries. Work is accepted only when five scenario walkthroughs produce an identified accountable owner, escalation clock, evidence standard, authority and remedy at every stage; regional teams validate feasibility; and metrics can be derived from sampled case records without manual reinterpretation. The design must preserve appeal independence and demonstrate that faster handling does not mean premature closure.

The client will provide de-identified case files, workflow data, current policies, vendor contracts, staffing and quality records, product roadmaps, regional counsel and protected access to frontline specialists. Consultants will not decide live cases, contact victims, prescribe policy outcomes, notify authorities or deploy detection models. Historical data remediation beyond the selected sample, product engineering and workforce implementation are excluded; access restrictions will be recorded as design constraints rather than filled with invented assumptions.

Why this is external work

Safety, product, legal and creator teams each defend legitimate but incomplete operating boundaries, while recent incidents have made internal facilitation politically charged. The company lacks a leader who has integrated high-harm investigations with monetisation and support at this scale. Independent architecture brings comparative case practice, protected challenge and a finite design that management can implement without outsourcing live accountability.

What you will own

• Reconstruct fifteen high-harm case journeys from first signal through decision, remedy, appeal and control learning.
• Define a severity taxonomy that distinguishes imminent harm, vulnerability, financial coercion, identity abuse and policy disagreement.
• Allocate decision rights across safety operations, specialists, payments, legal, product, regional teams and independent appeals.
• Model case demand, skill requirements, coverage hours and vendor limits without treating every report as interchangeable work.
• Test the model through five adversarial scenarios, recording delays, authority gaps, data barriers and unsafe shortcuts.
• Specify service measures that expose repeat harm, victim support, evidence quality, appeal outcomes and hidden queue transfers.
• Deliver the approved blueprint, mobilisation backlog, investment case and board decisions required before implementation begins.

Candidate qualifications

• Has designed trust-and-safety operations for a scaled social, creator, gaming or marketplace platform with material human risk.
• Can evidence integration of investigations, user support, payments, legal escalation, appeals and product controls around complex cases.
• Understands child-safety boundaries, evidence custody, regional escalation, trauma-informed support and procedural fairness.
• Has translated sensitive case evidence into staffing, workflow and decision architecture without exposing protected individuals.
• Can distinguish policy ambiguity from execution failure and explain where automation should never make the final judgment.
• Brings credible board communication and can preserve dissent among safety, growth, legal and product executives.

Non-negotiables

• Can lead protected evidence workshops in Los Angeles and complete four regional design sessions within five months.
• Will not review identifiable victim material outside approved environments or assume authority over live safety decisions.
• Has operated beyond content moderation queues; generic customer-service transformation experience is insufficient.
• Will document legal dependencies and evidence limits rather than presenting a falsely universal global workflow.

1. 49 words maximum. Which case handoff created the greatest hidden risk in a safety operation you redesigned, and how did you prove it?
2. 49 words maximum. What evidence would make you keep a high-harm decision outside an automated workflow?
3. 49 words maximum. Describe the smallest protected case sample you would use to test this operating model credibly.