Plaid head of AI Suddu Seshadri talks with Sophia Dew about training specialized models on transaction histories for financial risk and fraud detectionFraud detection identifies activity that may involve deception or unauthorized use by examining evidence and patterns..
He contrasts explaining a transaction with predicting risk from historical outcomes, and describes transformer-based modelsA transformer is a neural network architecture that uses attention to model relationships between elements of its input. and Plaid’s reported fraud-detection improvements. The interview presents those figures as company claims rather than independent results.
Seshadri also discusses personal financial agentsA personal AI agent uses a person's authorized goals, preferences, routines, context, and tools to provide continuing individualized assistance. and the need for trustworthy access controlsAn agent permission boundary defines and enforces which resources and actions an automated agent is allowed to use. as those systems take actions on behalf of users.
Watch on YouTube
