Offer a fixed-scope audit that compares a buyer's simulated scenarios with matched real-world driving logs, measures behavioral and sensor-level discrepancies, and prioritizes the gaps most likely to distort safety evaluations. Deliver a reproducible benchmark suite, realism scorecard, annotated failure cases, and a set of calibrated scenarios that can run inside the buyer's existing simulator. Begin as an expert-led service and productize recurring tests, metrics, and scenario-pack generation after repeated engagements reveal common requirements.

Autonomy developers are combining real-world data, synthetic generation, multimodal foundation models, and increasingly complex simulators, making independent validation more important and technically feasible. The concentration of hiring around simulation realism, foundational data quality, scene understanding, and trustworthy evaluation indicates an urgent capability gap.