Trust is the foundation of Mercor's marketplace. With hundreds of thousands of experts applying for jobs and over $XX million paid out every day, the integrity of who's on the platform and how they work directly determines the quality of the data we deliver to every lab and enterprise. Defending against adversarial, fast-moving threats is one of the highest-leverage problems at the company
You'll own the fraud detection and enforcement product and policies end to end: the multi-agent system that scores risk across expert lifecycle stages; the signals that feed it; and the enforcement actions. You'll also own the operational edge of trust and safety, designing the workflows reviewers live in, developing the governing policies, and closing the loop between what we catch in production and what we build next
Partner with engineering, data science, operations, legal, and GTM, and move fluidly between building operational systems, analyzing model changes and whether they actually improved business outcomes, and working directly with strategic customers
This is a technical, customer-facing role for someone who is energized by problems with no fixed playbook, who cares about both the system and the humans operating it, and who wants to own a product area where the stakes and the impact are unusually hig
Work backwards from the business: Connect Trust and Safety decisions to Mercor long term health, expert trust, and client trust, and use that to prioritize business outcomes
Own the Trust and Safety product end-to-end: Set the roadmap for Mercor's Trust and Safety systems and own the lifecycle from signal detection to enforcement action
Improve Detection quality: Work with data science and engineering to tune models, prompts, and signals against real caseloads driving down false positive rates while holding the line on false negatives and expand automated decisioning as accuracy improves
Build the Operations Tooling: Design the products and workflows that our operations analysts use every day. Orient them around the action being taken, keep evidence bars and audit trails clear, and make high-severity calls defensible
Get in the weeds: Write specs and PRDs, but also build dashboards, pull SQL, and make small PRs to unblock yourself and validate direction