Doctoral research on approximate computing for energy-efficient AI accelerators spanning edge and data-center platforms; emphasis on hardware–software co-design, low-precision algorithms, and circuit/architecture innovations to cut power and area while preserving accuracy.