Engineering Manager, Safety Processing

Discord Discord · Consumer · San Francisco, CA · Core Tech Engineering

Engineering Manager to lead a team building LLM pipelines for content review, enforcement systems, appeals paths, and investigator tooling for Discord's safety operations. The role focuses on automating safety reviews, improving appeals processes, and ensuring compliance.

What you'd actually do

  1. Lead three squads, Review, Enforce, and Tooling through their tech leads, setting direction across automated review, enforcement, and appeals
  2. Push automation toward nearly all review decisions without giving up precision or recall against a human baseline, and get engineering out of the loop so the operations team can launch and tune workflows themselves
  3. Rebuild appeals: one path for every eligible appeal regardless of what triggered the original action, review tooling that scores decision quality rather than only flagging that two reviewers disagreed, and checks that catch mistakes before they land on good users
  4. Own the tooling our investigators depend on, cutting complex investigations from hours to minutes, and make enforcement traceable enough that anyone can see how a user reached their current state
  5. Carry regulatory and child-safety compliance work alongside the roadmap, set the engineering bar with your leads, and hire and grow the team

Skills

Required

  • 5+ years as a software engineer with a solid backend and distributed systems background
  • 2+ years managing engineers
  • Shipped LLM systems in production where the output is a decision the business acts on
  • Experience working alongside operations or human-in-the-loop teams
  • Excellent communication and cross-functional collaboration
  • Recruiter and coach

Nice to have

  • A strong passion for Discord and/or gaming
  • Trust and safety, fraud, integrity, or abuse engineering at a consumer platform
  • Building internal tools where operations or review teams are the primary user
  • Consolidating fragmented legacy systems onto a single platform, or managing outsourced operations teams and vendors

What the JD emphasized

  • LLM systems in production where the output is a decision the business acts on
  • opinions about evaluation, ground-truth data, quality regressions, and where cost and latency start to bite
  • Legal deadlines, serious incidents, and regulator attention come with this job
  • regulatory and child-safety compliance

Other signals

  • LLM pipelines for content review
  • automation of safety reviews
  • appeals path for users
  • tooling for investigators
  • regulatory and child-safety compliance