AI & ML interests

Shrink model size without compromising performance.