Over the last decades, implantable neural interfaces have been extensively explored and effectively deployed to address neurological and mental health disorders. The existing solutions present several limitations. Firstly, the physical size of the implantable device remains bulky, primarily due to the inclusion of batteries. Second, the presence of cables significantly hamper biomechanical compliance, restrict the maximum number of channels and make the surgical procedure inherently complex. Wireless power transfer and CMOS miniaturisation are key factors in addressing those limitations. Up to date, the smallest wireless neurostimulator relies on 3-coil inductive link and delivers ±25 µA in a total area of 650x650 µm2. However, the system cannot continuously operate complying within safety standards, the maximum current is limited, and the manual electrode integration reduces the total yield. Hence, an imperative demand exists for innovative solutions enabling system miniaturisation by safely exploiting wireless power transfer and communication while allowing sufficient charge injection capabilities. In this thesis, a miniaturised, wireless, and distributed neural interface is ideated and developed envisioning a large-scale cortical prosthesis with thousands of free-standing and individually addressable implants. A novel frequency-switching inductive link is proposed to overcome safety and performance limitations of traditional systems. Preliminary results show an improved efficiency and delivered power respectively of two orders of magnitudes and more than six decades compared with state-of-the-art 3-coil links while working with 1024 receivers with the up-to-date smallest size of 200x200 µm2. As a significant advancement, the introduced "Neural Dot" envisions a fully integrated monolithic chip including the receiver coil as well as the analog and digital functions. All the CMOS blocks are individually tested, showing the smallest power manager and the smallest biphasic current-controlled stimulator capable of delivering train of bursts of ±40 µA with a limited harvested power budget at the rectifier input. Circuit design is co-optimized with the electrode/brain interface, engineering the Pt-black coating material and studying the effect of the bioimpedance changes over time by proposing a non-stationary model for robust circuit co-optimization. A novel electrode microfabrication technique relying on CMOS-compatible post-processing at the wafer level is proposed to address the traditional challenges in integrating the electrodes with miniaturised chips. This "foundry-to-surgeon" approach envisions a multi-shank and multi-channel scalable and large-scale neural interface. All the investigations were carried out toward a large-scale cortical visual prosthesis, for which every single CMOS stimulating unit corresponds to a phosphene perceived in a precise spatial location of the blind visual field. In vivo studies with non-human primates were conducted to derive the main trade-off among the stimulation parameters for developing and co-optimizing the miniaturised CMOS unit, representing an effort well beyond the present state-of-the-art for both area consumption and power budget. The preliminary results motivate further investigation and development as the proposed solutions are promisingly addressing several critical issues commonly encountered today, thus opening new frontiers for treating neurodegenerative and mental health disorders.