Research Validation Expert (Synthetic Respondents / Digital Twins)

We are a research-data company that has built a new way to generate survey responses using AI grounded in the real, observed behavior of actual individuals rather than invented personas. We have benchmarked it ourselves against published sources and the early results are strong. Now we want experienced, independent people to pressure-test the method and tell us where it holds up and where it breaks. We are looking to build a small bench of senior contractors for one or more short engagements. This is not a data-entry or survey-fielding gig. We want methodological judgment. What the work involves: Reviewing our validation approach and benchmark design, and telling us honestly where it is weak Designing or running independent comparisons of our AI-generated data against trusted sources or fresh human samples Going beyond topline accuracy to assess subgroup estimates and whether the relationships between variables are preserved Writing up findings clearly enough that a skeptical buyer or partner would trust them Who we are looking for: 10+ years in survey research, marketing science, polling, or applied statistics Hands-on experience with validation, benchmarking, weighting, and error measurement (MAE, MSE, calibration, significance testing done sensibly) Comfort with subgroup analysis and multivariate relationships, not just marginals A reputation for rigor and a willingness to deliver an unflattering result if that is what the data shows Independent, meaning not currently tied to a competing synthetic-data or panel vendor Nice to have: Familiarity with synthetic respondents, digital twins, or LLM-based survey simulation Prior work with sources like Pew, Gallup, Nielsen, MRI-Simmons, Census, or election data Past experience designing validation studies for the ARF, ESOMAR, MRS, or similar