Fantix is an AI platform for business data science. We train robust machine learning models without pooling user-level data, thus protecting consumer privacy and business confidentiality. Our models include: Yellowcake™️: a universal, generative model for consumer data enrichment that is capable of predicting hundreds of attributes of a consumer without knowing their identity. It is trained on a large variety of datasets of payment transactions (merchant and SKU-level), mobility data and demographics, and is over 90% accurate. Fusion™️: a federated audience builder that creates lookalike and advanced targeting audiences for advertisers. Leveraging advanced deep learning techniques and federated data from multiple partners, this system ensures enhanced precision and privacy in audience targeting. Supernova™️: a model that generates synthetic data sets, designed to mimic real world data without being PII. Financial institutions use these privacy-proof datasets to research stocks, product teams use them for market research. Trained on credit card transaction data and mobility data, these models are available mapped by merchant and SKU. LightID™️: a cookieless ID resolution model that identifies up to 90% of site visitors (it can positively identify unlogged traffic).