Program Manager, Human Data

OpenAI OpenAI · AI Frontier · San Francisco, CA · Research

The Human Data team at OpenAI is looking for a Program Manager to design and execute programs for collecting high-quality human feedback data. This role involves partnering with research, operations, and engineering teams, managing external vendors and AI trainers, and ensuring the successful completion of data campaigns. The goal is to use this data to train and evaluate AI models, contributing to the development of safe and effective AI systems.

What you'd actually do

  1. Work closely with external vendors, trainers and internal researchers to collect, review, and deliver high-quality data
  2. Gather requirements, write instructions, define success criteria, and calibrate the AI trainers
  3. Use internal tooling to assess labeled data and provide feedback to AI trainers
  4. Think critically and share recommendations on tooling and process improvements, optimizing for quality, throughput, and AI trainer experience

Skills

Required

  • Program management
  • Vendor management
  • Data collection and review
  • Instruction writing
  • Success criteria definition
  • Calibration of trainers
  • Data assessment
  • Process improvement
  • Communication
  • Organization

Nice to have

  • Interest or background in AI, LLMs, Agents
  • Hands-on experience with data analytics
  • Experience in dynamic or start-up environments

What the JD emphasized

  • high-quality data
  • AI trainers
  • labeled data
  • tooling and process improvements

Other signals

  • human feedback into reliable signals for training and evaluation
  • design and run end-to-end programs that capture the depth of human intent
  • scalable synthetic data generation
  • translate these signals into training datasets, novel evaluations
  • partner with our research teams, operations and engineering to execute complex programs for collecting high-quality data
  • key interface between our external vendors and AI trainers
  • ensuring human data campaigns are successfully completed
  • train safe models that will land in the real world
  • assess labeled data and provide feedback to AI trainers
  • optimizing for quality, throughput, and AI trainer experience