Model Compression & Edge AI

The question is not “how small can we make it” — it is “what do we lose, and does it matter for the deployment we care about?” A map of the software foundations, hardware physics, and deployment constraints where compression stops being optimisation and becomes existential.