When
Where
Webinar Materials
Presentation Slides
About the Webinar
You've got a ton of data to analyze and you need to track your project end-to-end – from data pre-processing, through model optimization and evaluation, to deployment and analysis. And, perhaps you also have scores of simulations for which their outputs, corresponding parameters, package dependencies, and source code need to be tracked and stored all in one place. Sound familiar? Join us for a demonstration of MLFlow, a platform-agnostic and collaboration-friendly machine learning lifecycle manager that enables seamless teamwork where each member's individual progress can be tracked and reproduced. Presenters Ariyan Zarei and Artin Majdi will demonstrate how they've applied MLFlow to a PhytoOracle modeling project using ag-based remote sensing data to enable a robust, efficient, and reproducible ML pipeline that helps you get the job done and done right.
Watch an earlier webinar about the PhytoOracle project.
What You'll Learn
- What MLFlow is and does
- The advantages and disadvantages of using MLFlow
- How MLFlow has been applied to a workflow project using ag-based remote sensing data
Suggested Level of Expertise
- Familiarity with Machine Learning concepts, terminology, techniques, and principles
- Using command line for data analysis
- Understanding model packaging, deployment, and versioning for project management
About the Presenters
Ariyan Zarei is a key member of the PhytoOracle "digital ag" team and is currently pursuing his PhD in Computer Science at the University of Arizona. For PhytoOracle, Ariyan is helping design the machine learning, computer vision, and statistical models for geo-correction and stitching of high-resolution RGB image data.
Artin Majdi is a Graduate Assistant at the University of Arizona's Data Science Institute and a doctoral candidate in the Electrical and Computer Engineering department. His areas of research include image analysis and processing, ML data and model uncertainty, advanced neural networks, and ML lifecycle management.