Wells Fargo pre-trained its own foundation model on billions of transactions to beat XGBoost at fraud detection
AI Engineer · youtube · 2026-10-23
Freddy Lecue of Wells Fargo explains how the bank pre-trained a transformer from scratch on billions of customer transactions, learning to predict the next transaction. One backbone is fine-tuned for fraud detection, credit line decisions, AML and marketing. Key challenge: tokenizing tabular data (amounts, timing, merchants, zip codes). On fraud it caught more fraudulent transactions with fewer false alarms than the production XGBoost model, while staying fast enough for real-time card approvals.
More from Companies & People
- Jensen Huang visits Astana, joins Kazakhstan's AI minister to discuss nation's AI future — nvidia · 2026-10-23
- Microsoft's FrontierTour: AI governance becomes everyone's problem as agents take action — donasarkar · 2026-10-23
- Why Netflix keeps losing to YouTube: UGC platforms professionalize easier than pro platforms go casual — ZeroStateReflex · 2026-10-23
- Stanford releases free 520-page robot autonomy course book and lectures — lukas_m_ziegler · 2026-10-23
- xAI reportedly launching 'Friends of Grok' program with Ultra perks and build credits — nima_owji · 2026-10-23
- Founders tour Greater Bay Area: a year of hardware prototyping done in 3 months — DominiqueCAPaul · 2026-10-23