A CRO platform combining Taguchi factorial design for variant selection, genetic algorithms for crossover and mutation, and Bayesian sequential testing with Thompson Sampling for live winner-detection without fixed sample sizes. An LLM generates variants grounded in a copywriter's 15-year body of agency craft. A neuroscience collaborator is adding a neural-network layer modelling the emotional-state transitions buyers flick between during a purchase decision, feeding the variant generator. Stack: Python, FastAPI, NumPy, SciPy, PostgreSQL, Anthropic Claude, Next.js.