Week 4, Scientific paper proposal

[Arbmani]

Assignment Proposal

Title

Machine Learning Operations (MLOps): Overview, Definition, and Architecture

Names and KTH ID

• Oscar Arbman (oarbman@kth.se)

• Dania Sami (dsami@kth.se)

Deadline

• Week 4

• Topic: MLOps/AIOps/LLMOps

Category

• Scientific paper

Description

This paper first asks: "What is MLOps?", and by using a literature review, tool review, and eight expert interviews, they provide for MLOps " an aggregated overview of the necessary principles, components, and roles, as well as the associated architecture and workflows, and a clear definition of what MLOps is, and outline unresolved issues in MLOps. According to the paper, it: "provides guidance for ML researchers and practitioners who want to automate and operate their ML products with a designated set of technologies". They also determine key foundations of MLOps, extract the essential building blocks, emphasize the positions required for effective MLOps adoption, and formulate a general architecture for ML system design. According to the authors of the paper: "These insights can assist in allowing more proofs of concept to make it into production by having fewer errors in the system’s design and, finally, enabling more robust predictions in real-world environments."

• Link to article: https://ieeexplore.ieee.org/abstract/document/10081336

Relevance

This paper is relevant to MLOps/AIOps/LLMOps, because it outlines key challenges to MLOps such as "Organizational challenges", "ML System challenges", and "Operational challenges", and provides a framework for MLOps in terms of principles, roles, and architectures that, according to the authors "can assist in allowing more proofs of concept to make it into production".

[frankreyesgarcia]

Hi, the proposal is not appropriate for that week. It is more appropriate for the CI and Automated Testing week. I recommend looking for another proposal.

[Arbmani]

Hi, the proposal is not appropriate for that week. It is more appropriate for the CI and Automated Testing week. I recommend looking for another proposal.

Since model validation is part of MLOps, could it not be argued that researching whether AI can perform effective automated testing, such as unit tests, might allow AI to validate machine learning models or make changes to pass tests? This could enable it to improve itself iteratively, and therefore, this topic should be relevant to MLOps.

[frankreyesgarcia]

In this case, you are not testing the models; rather, you are using them to generate tests. That is why it does not apply to the week you choose.

[Arbmani]

We have changed the proposed paper

[Arbmani]

Nvm I think this paper is not in the list of allowed sources, so I'm closing the pull request