Abstract

There’s a lot more to Machine Learning (ML) than just utilizing a ML algorithm and building a model out of it. In real world, production ML systems are large ecosystems of which building the model is just a small part. The real challenge comes in when these models must be deployed and

maintained in production to generate real business value. Some of the challenges are building ETL pipeline, monitoring data/model drift, automated model training/deployment, feature engineering, model metadata management etc. Due to these challenges, less than 20% of the total models build make it to production and offer some business value. Building reliable, scalable, and reusable machine learning pipelines is a demanding task and it takes immense

efforts to build and maintain them. As a machine learning engineer, I will depict how to transit your experimental ML models to production by building a production ready ML pipeline in simple steps using popular MLOps framework.

Bio.

Satyajit Padhy received his Ph.D. degree with the Institute of Information Systems and Applications, National Tsing Hua University, Taiwan. His PhD research focused on resource management algorithms in Network Function Virtualization. His industry experience started with Industrial Technology Research institute where he was the lead engineer in developing an Auto Machine Learning system & building ML model for edge devices. He joined Micron as a senior ML engineer in 2021 where he is responsible in building and maintaining production ML pipelines.

All faculty and students are welcome to join.