How AI infrastructure differs from common infrastructure [ukr]

For the past three years, I have worked at LetsEnhance/Claid, an AI startup that enhances image quality and creates new backgrounds for customer products. While the infrastructure is always the same, it varies: large models, significant latency, and unreliable autoscaling. In this speech, I will discuss our perspective on AI infrastructure and our ideas for its improvement.

The topic is for mid to senior-level engineers who are either already in AI or interested in understanding how it differs from traditional infrastructure.