How to use open-source tools to serve an AI/ML model on private infrastructure
AI Services like ChatGPT and Dall-E are already popular, and their advantages are undeniable, but there’s still a general concern around the topic of data privacy.
The purpose is to allow users that value their privacy to access the same kind of services that are provided over the network by companies like Google and OpenAI,but running on their own infrastructure and with no need for their data to travel across the net. This talk will onboard users on how to deploy and serve, or even handle the training lifecycle of, ML models with just a few tools and a base project.
Using k8s and cdk8s for infrastructure, users can easily setup and configure their own AI powered services, making it possible to deploy to a private cloud service, on prem infrastructure, custom hardware (raspberry pi) or AI ready devices (https://www.nvidia.com/en-us/data-center/a100/)