Data assimilation (DA) shares many mathematical similarities with methods in artificial intelligence and machine learning (AI/ML). New methods are being developed to leverage these connections to improve capabilities in coupled data assimilation (CDA) for the initialization of coupled forecast models. Such coupled models are being used by operational prediction centers for forecast applications ranging from medium-range weather forecasts to subseasonal-to-seasonal (S2S) prediction. Advances by the integrated AI/ML & DA research group at CIRES and the NOAA Physical Sciences Laboratory will be discussed.