Rebuild an AI evaluation console for support operations

Khan AI Lab is replacing an early internal console used to review language-model outputs for customer support teams. The present tool is difficult to extend and does not make disagreement between reviewers, source data and model results easy to inspect. The new console should be delivered through a phased roadmap with review queues, dataset versioning, evaluation criteria and auditable result views. Payment will be released at the agreed delivery point. Acceptance requires an evaluator to compare outputs against a labelled case, record a judgement, and see the updated quality trend.