OpenAI DeepMind & Anthropic Employees - Human Data Sourcing & Buying

Expert Sourcing Brief — Human Data for AI Training Objective I’m looking to speak with current or recent employees from OpenAI, Google DeepMind, or Anthropic who have direct knowledge of how frontier AI labs source, evaluate, and purchase human-generated data for model training Expert Sourcing Brief — Human Data for AI Training Objective I’m looking to speak with current or recent employees from OpenAI, Google DeepMind, or Anthropic who have direct knowledge of how frontier AI labs source, evaluate, and purchase human-generated data for model training, post-training, and evaluations. The goal is to better understand how the human data / data authoring industry is evolving and what leading AI labs are actively looking for from external data providers today. Ideal Expert Profile The expert should have worked directly on, or closely with, areas such as: Human data sourcing / data partnerships Post-training data RLHF / RLAIF / reinforcement learning data Expert data / domain-expert sourcing Model evaluations and benchmark creation Data operations or data quality RL environments / agent training environments Vendor management for training-data providers Partnerships with companies such as Scale AI, Surge AI, Mercor, Handshake, Turing, Micro1, etc. Preferred seniority: Manager, Lead, Director, Researcher, Product Manager, Data/Operations Lead, or Partnerships Lead with meaningful visibility into purchasing decisions or data strategy. Companies Priority: OpenAI, Anthropic, Google DeepMind Secondary: Meta AI, Microsoft AI, xAI, Amazon AGI, Mistral, Cohere, Thinking Machines, or other frontier-model labs. Key Questions I Want to Understand What kinds of human data are AI labs buying today? Coding Finance Legal STEM General knowledge work Multimodal Agentic / long-horizon tasks RL environments Evals and benchmarks What is becoming more important over the next 12–24 months? Human preference data? Expert-authored tasks? Rubrics and verifiers? Real-world enterprise data? Agent trajectories? Screen recordings? Synthetic environments? Domain-specific evaluation sets? How do labs choose between data vendors? Quality Access to experts Speed Scale Price Data diversity Fraud prevention Technology/platform capabilities Ability to build RL environments Ability to demonstrate model uplift Where are the biggest current bottlenecks? Finding qualified experts Data quality Creating difficult enough tasks Writing good rubrics/verifiers Building realistic environments Reviewing data Fraud / identity verification Turnaround time What expert profiles are most in demand? I’d like to understand which professions and levels of seniority labs currently struggle most to source—for example: Software engineers Investment bankers / PE professionals Lawyers Doctors Scientists Consultants Accountants Designers Other specialized professionals How is the market changing? Is “data labeling” becoming “data authoring”? Will demand for humans increase or decrease as models improve? How much can synthetic data replace human-created data? Are labs consolidating spend with a few large providers? Where is there still room for new or specialized suppliers? Important I am not looking for confidential, proprietary, or non-public information. I’m interested in the expert’s industry perspective, experience, and general understanding of current market needs and trends.