Abstract
On-demand generative AI inference lowers many of the traditional barriers for deploying bigger models: from pre-training and alignment to inference infrastructure.
Out of all of the available models and architectures, which ones should we choose for a specific use case?
This ultimately depends on the available data and what experiments can be feasible or practical.
We will discuss some concrete use cases with varying generalizability expectations. For some it is enough to just use OpenAI and not worry while in others proper statistical significance is warranted.

Bio
Georgi is the CEO of Crisp Labs, an Applied AI Advisory. His research background is in Reinforcement Learning and has 10+ years of data science experience across quantitative finance, healthcare and manufacturing.

Bookings