Add the GreatAI deployment framework to the list

[schmelczer]

Short description

GreatAI: An easy-to-adopt framework for robust end-to-end AI deployments investigates a way to increase the adoption rate of ML deployment libraries (and hence the overall quality of deployed AI services) by also focusing on ease of use. One of its outcomes is a concrete library — called GreatAI — implementing this philosophy.

Why is it relevant?

The research presents 33 AI/ML deployment best practices, the difficulties coming from trying to implement them, and ways to overcome these challenges. The GreatAI library provides streamlined, automated implementations for 17 of these SE-ML best practices.

What is the lesson learned from reading the article?

Facilitating the adoption of AI deployment best practices is viable by finding less complex API designs. Additionally, it shows that many best practices can be given automated implementations.

How does it add new insights over the articles already listed?

• The research component is novel because no previous work has tried simultaneously balancing both ease of adoption and deployment quality.

• The GreatAI library tries to mirror the simplicity of modern AI libraries such as Hugging Face transformers and bring their simplicity into the field of SE-ML. It has fewer features than Seldon Core, AWS SageMaker, or TensorFlow Extended but using it only requires a couple of lines of code. Thus, it might strike the right balance for ML deployments that otherwise wouldn't have prioritised deployment best practices.