MLCask

A Git-like end-to-end ML life-cycle management system

View our publication in the 2021 IEEE International Conference on Data Engineering (ICDE):

MLCask: Efficient Management of Component Evolution in Collaborative Data Analytics Pipelines

Research Highlights

MLCask is a Git-like end-to-end ML life-cycle management system. In real-world machine learning (ML) applications, maintaining an ML pipeline in a collaborative environment is significant and challenging. The costs of frequent retraining and asynchronous component update by different users need to be taken into consideration. MLCask supports both linear and non-linear version control semantics for efficient management of ML pipelines.

Our Research

Oct 8, 2021

MLCask Overview

Project MLCask is a Git-like end-to-end ML life-cycle management system. MLCask supports both linear and non-linear version control semantics for efficient management of ML pipelines. The system abstracts an ML life-cycle with two key concepts: component and pipeline. A component refers to any computational units in the ML pipeline, while a pipeline is the minimal unit that represents an ML task.

Oct 7, 2021

Example Applications

MLCask supports both linear and non-linear version control semantics for ML pipelines of different applications. Here, let us consider the use case of readmission. Linear Version Control Semantics We use Directed Acyclic Graph (DAG) to formulate an ML pipeline.