Paper Title
A Transformer Based Model for Continuous Neural Financial Forensics

Abstract
Abstract - This paper presents a deep learning model for Continuous Financial Forensics. All the large financial institutions of the world are finding it very difficult to predict and control the different type of financial crime, especially if the type of the attack is new and the time of attack or crime is unknown. The extent of the unknown ranges across all the types of financial crime starting from terrorist financing to insider trading, money laundering and a host of other financial crime types. Financial institutions are having access to a huge dataset which includes transactional as well as customer data along with their social data – however there is no established model for Neural Financial Forensics. The proposed Deep Learning model uses Transformer Architecture on modern language models and extends on the modern engineering concepts of Agile, DevOps and MLOps and proposes a framework which can be used by financial institutions to predict and control financial crime. The model leverages on the principles of Explainable AI (XAI) and recommends a Zero Shot Learning (ZSL) approach to detection. Keywords - Neural Network, Deep Learning, Financial Forensics, Language Model, DevOps, MLOps, XAI