1. Project Overview

Prophecis All-in-One Machine Learning Platform

Prophecis is an all-in-one machine learning platform independently developed by WeBank. It integrates multiple open-source machine learning frameworks, supports multi-tenant management of machine learning computing clusters, and provides full-stack container deployment and management services for production environments.

2. Open Source License

Licensed under Apache-2.0

3. Screenshots

• Key Features

Figure: Prophecis Features

4. Feature Overview

• Full-Lifecycle Machine Learning Experience
  : Prophecis MLFlow can be integrated into DataSphere Studio workflows through AppJoint, supporting the entire machine learning pipeline, from data uploading, data preprocessing, and feature engineering to model training, model evaluation, and model deployment;
  

Figure 3: Prophecis Integration with DSS

• One-Click Model Deployment Service
  : Prophecis MF supports one-click publishing of trained models generated by Prophecis Machine Learning Flow and Prophecis MLLabis as RESTful API or RPC interfaces, enabling seamless integration of models into business applications;
• Comprehensive Management Platform for Machine Learning Application Deployment, Operations, and Experimentation
  : Customized based on open-source community solutions, it provides a complete, reliable, and highly flexible set of enterprise-grade management tools for machine learning application deployment, monitoring, service governance, log collection, and querying. It enables comprehensive control over machine learning applications and meets all enterprise requirements for running machine learning applications in production environments.

5. Tech Stack

Architecture

• Overall architecture

• Figure: Prophecis Overall Architecture
• Prophecis mainly consists of 5 key services:
• Prophecis Machine Learning Flow
  : A distributed machine learning modeling tool that supports both standalone and distributed model training. It supports multiple machine learning frameworks, including TensorFlow, PyTorch, and XGBoost, and provides a complete pipeline from machine learning modeling to deployment;
• Prophecis MLLabis
  : A machine learning development and exploration tool that provides development and exploration services. It is an online IDE based on JupyterLab, supports machine learning modeling tasks on GPU and Hadoop clusters, supports Python, R, and Julia, and integrates various plugins such as Debug and TensorBoard;
• Prophecis Model Factory
  : A machine learning model factory that provides services for model storage, model deployment testing, and model management;
• Prophecis Data Factory
  : A machine learning data factory that provides feature engineering tools, data annotation tools, and resource management services;
• Prophecis Application Factory
  : A machine learning application factory jointly developed by WeBank's Big Data Platform team and AI department. Customized based on QingCloud's open-source KubeSphere, it provides CI/CD and DevOps tools, as well as monitoring and alerting capabilities for GPU clusters.