Implantable neural interfaces promise transformative therapies for brain disorders unresponsive to conventional treatments. Despite significant advances in neural interface microsystems over the past decade, the small number of recording and stimulation channels and limited embedded processing in the existing clinical-grade technologies remain a barrier to their therapeutic potential. In this talk, I will present an overview of our research on the integration of modern signal processing and machine learning techniques in neural interface System-on-Chips (SoCs) for epilepsy, movement, and mental disorders, and for the next-generation implantable brain-machine interfaces (BMIs). Our goal is to develop a new generation of intelligent, highly integrated CMOS-based devices to address a broad spectrum of brain disorders in the future.