Architecture that tracks data distribution, data generation and calculate impact of data amplification campaigns on a GPU-hosted zoo of specialist LLM models, to test, learn and optimise faster. 7.