This work presents a novel approach to enable scalable and energy-efficient AI for machine learning (ML) and edge applications. As the demand for AI continues to grow across various industries, there is an ever-growing need for higher computational power and area-efficient computing methods. Traditional digital computing methods employed by the industries, such as using digital accelerators, are highly power-intensive and area-inefficient, posing a significant challenge for various design applications. Such systems also significantly impact the environment in terms of enormous carbon emissions footprints and is unsustainable in the long run. This research argues that analog AI hardware and computing methodologies offer a unique alternative that overcomes these limitations and offers unmatched power density, performance, and area benefits compared to digital methods employed today. To date, the power density and performance benefits of analog designs remain unmatched by their digital counterparts. However, analog technology has historically struggled to be used in large-scale systems due to several challenges associated with it. This research aimed to overcome these challenges and demonstrate the potential of analog computing for building large-scale AI systems. In this regard, this work also showcases India's first Analog AI Accelerator chipset, called ARYABHAT (Analog Reconfigurable Technology And Bias-scalable Hardware for AI Tasks), built on this novel in-house technology and simultaneously presents a complete analog compute ecosystem to sustain this technology for various AI applications.