The radio access network (RAN) is moving towards open interfaces that offer wireless applications rich opportunities for customization and optimization. In 5G and beyond, the use of a large number of antenna elements, referred to as Massive MIMO, is envisioned to be the key physical layer enabling technology to transform wireless access into a high-throughput, low-latency multi-user medium where interference can be mitigated based on beamforming techniques. When the control of these large number of antenna elements are exposed via open interfaces, the RAN becomes a platform for the next-generation learning-based intelligent algorithms to deliver self-optimizing network access and connectivity.