Reusable MLOps: Reusable Deployment, Reusable Infrastructure and Hot-Swappable Machine Learning models and services

Link to our paper: https://arxiv.org/abs/2403.00787

This paper 'Reusable MLOps: Reusable Deployment, Reusable Infrastructure and Hot-Swappable Machine Learning models and services' introduces a new sustainable concept in the field of AI/ML operations is introduced - called Reusable MLOps - where the existing deployment and infrastructure is reused to serve new models by hot-swapping them without tearing down the infrastructure or the microservice, thus achieving reusable deployment and operations for AI/ML models while still having continuously trained models in production.

Abstract: Although Machine Learning model building has become increasingly accessible due to a plethora of tools, libraries and algorithms being available freely, easy operationalization of these models is still a problem. It requires considerable expertise in data engineering, software development, cloud and DevOps. It also requires planning, agreement, and vision of how the model is going to be used by the business applications once it is in production, how it is going to be continuously trained on fresh incoming data, and how and when a newer model would replace an existing model. This leads to developers and data scientists working in silos and making suboptimal decisions. It also leads to wasted time and effort. We introduce the Acumos AI platform we developed and we demonstrate some unique novel capabilities that the Acumos model runner possesses, that can help solve the above problems. We introduce a new sustainable concept in the field of AI/ML operations - called Reusable MLOps - where we reuse the existing deployment and infrastructure to serve new models by hot-swapping them without tearing down the infrastructure or the microservice, thus achieving reusable deployment and operations for AI/ML models while still having continuously trained models in production.

Citation:

Panchal, D., Verma, P., Baran, I., Musgrove, D., & Lu, D. (2024). Reusable MLOps: Reusable Deployment, Reusable Infrastructure and Hot-Swappable Machine Learning models and services. ArXiv, abs/2403.00787.

Author Bio:

Deven Panchal is an AI Leader currently with AT&T Labs Research, USA. He also serves as the Chief Innovation and IP/Patents Officer of the AT&T Innovation Network and is a Senior Member of the IEEE.

He has led the launch of 4 products into Open Source, being used worldwide today. He has delivered products and platforms projected to save ~1 billion$, enable Network Services creation, Network Management, Service Management and Orchestration and to Improve Customer service. He has helped build the world’s first AI/ML marketplace, was involved with AT&T’s first 5G VRAN/CRAN trial and delivered solutions that today power AT&T’s network as well as other telco networks. His work has benefited millions of people around the world, received multiple awards and has been featured on Forbes, TechCrunch, Light Reading, TelecomTV etc.

In the past, he has served as a Research Scientist at SAMEER, IIT Bombay building India's first indigenous LINAC and MLC for Cancer Radiotherapy.

Deven Panchal has been the Owner for several subprojects within Acumos, and has led global teams in the Design and Development of these components. He has written several papers on Acumos, and has contributed heavily to the official Acumos documentation. Additionally, he has been a Top Code Contributor and Peer reviewer for Acumos.

Feel free to reach out to him or use/cite his work.