Predicting grain formation during alloy solidification is of great importance in additive manufacturing. Numerical simulations require fine spatial and temporal discretization that can be computationally expensive. In this talk, I will discuss GrainGNN, an efficient and accurate reduced-order model for epitaxial grain growth in additive manufacturing conditions. GrainGNN is a sequence-to-sequence long-short-term-memory deep graph neural network that evolves the dynamics of manually crafted features. In GrainGNN, we use a dynamic graph to represent interface motion and topological changes due to grain coarsening. We use a reduced representation of the microstructure using hand-crafted features; we combine pattern finding and altering graph algorithms with two neural networks, classifier (for topological changes) and a regressor (for interface motion). Both networks have an encoder-decoder architecture; the encoder has a multi-layer transformer long-short-term-memory architecture; the decoder is a single layer perceptron. We present results in which GrainGNN can be orders of magnitude faster than phase field simulations, while delivering 5%–15% pointwise error. This speedup includes the cost of the phase field simulations for generating training data. GrainGNN enables predictive simulations and uncertainty quantification of grain microstructure for situations not previously possible.