Software Engineer, Integrity Foundations - London

OpenAI OpenAI · AI Frontier · London, United Kingdom · Applied AI

Software Engineer role focused on building scaled foundational systems for detecting, tracking, and enforcing harm on OpenAI's platforms, leveraging advanced AI technologies to automate and augment these processes. The role involves collaboration with various experts and staying ahead of adversarial threats.

What you'd actually do

  1. Design and build scaled foundational systems used in the detection, tracking, and enforcement of harm using our technologies.
  2. Work closely with and learn from technical and non-technical experts on harms we are observing, and that stem from our platforms, to inform your designs and implementation.
  3. Leverage OpenAI’s most advanced technologies to automate and augment our abilities to detect and reason about complex harms quickly, accurately, and with minimum human intervention.
  4. Collaborate with policy, trust and safety operations, legal, investigations, and harm-specialised engineers and data scientists to holistically combat abusive actors and customers using OpenAI’s technology.
  5. Stay abreast of the latest techniques and tools to stay several steps ahead of determined and well resourced adversaries.

Skills

Required

  • backend software engineering
  • data systems
  • trust and safety analysis
  • investigation
  • operations

Nice to have

  • Machine Learning techniques

What the JD emphasized

  • at least 5 years of software engineering experience in backend and data systems
  • at least 2 years experience in trust and safety analysis, investigation, and/or operations

Other signals

  • detecting and responding to bad actors and harm on OpenAI’s platforms
  • architect next-generation systems that will support, leverage, and scale the work of experts in harms committed with our technology
  • Leverage OpenAI’s most advanced technologies to automate and augment our abilities to detect and reason about complex harms quickly, accurately, and with minimum human intervention