The objective of this project is to use high-speed thermal images of AM processes to drive physics informed machine learning (PIML) approaches that will help discover melt-pool dynamics. Predicting and learning the underlying melt pool dynamics in additive manufacturing processes is critical to realize flawless parts. Current process monitoring tools provide little insight to complex heat transfer and phase change processes in the melt pool. Simulation tools like Flow-3D offer impressive physics-based packages to understand these processes, but are computationally expensive and demand validation over a broad range of process parameters. PIML can create tools that explore knowledge gaps in the underlying physics and more efficiently simulate the process, thus enabling real time qualification and eventually process control.