Andrew Green, Managing Director and lead XVA Quant, Scotiabank Quantitative Analysts in banks historically built valuation models using analytic approaches coupled with numerical techniques like Monte Carlo, tree models and PDE solvers. Today, Quants are exploring machine and deep learning techniques as adjuncts and alternatives to traditional models. In this presentation I explore one use case in which Deep Neural Networks can be combined with Monte Carlo methods to give a high performance and flexible calculation platform for XVA. XVAs or derivative valuation adjustments are portfolio level modifications to the value of derivative transactions due to effects such as counterparty credit risk (CVA) and funding costs (FVA). Calculating XVAs is computationally demanding and mathematically complex, while the XVAs themselves are often a material component of bank balance sheets. Here deep learning is used to approximate derivative models, providing a fast, parallel valuations that can be used as a component to accelerate the XVA calculation itself.