🎥 Recorded live at the MLOps World | GenAI Summit 2025 — Austin, TX (October 8, 2025) Session Title: Fake Data, Real Power: Crafting Synthetic Transactions for Bulletproof AI Speaker: Bhavana Sajja, Senior Machine Learning Engineer, Expedia Inc. Talk Track: Data Engineering in an LLM Era Abstract: Can your AI models stay powerful without exposing sensitive customer data? In this hands-on talk, Bhavana Sajja from Expedia shows how to generate high-quality synthetic transaction data that behaves like real data—without revealing a single private detail. You’ll learn how to overcome the toughest challenges in financial and transactional datasets: mixed data types, rare events like fraud, and complex feature dependencies. Bhavana walks through four leading generative approaches—GANs, TVAEs, TabularARGNs, and GPT-based methods—explaining how each can be applied to build secure, realistic, and effective AI datasets. Using a practical case study, she demonstrates how to clean, encode, and model data to train fraud detection systems that remain accurate, private, and compliant. This talk is your blueprint for balancing privacy, utility, and innovation—using synthetic data to unlock insights without risk. What you’ll learn: • Why high-quality synthetic data is key to privacy-preserving AI • How to generate realistic transactional data with GANs, TVAEs, and GPT-based models • The trade-offs between data utility and privacy protection • How to evaluate the performance of models trained on synthetic data • Real-world applications for synthetic data in finance, healthcare, and IoT.