1 Introduction
You are a Data Scientist struggling with data, code, and models in your projects.
You are an ML Engineer who has trouble replicating pipelines and monitoring models in production.
You are Head of a Data Science team, who has trouble shipping quickly models to production.
You are a CTO who has difficulty to move Python code into large-scale production and manage the DevOps workload on an AI project.
🚀 Then, you are a potential user of Craft AI’s MLOps platform, which aims to accelerate the deployment and improve the management of your Machine Learning models in production.
1.1 What is Craft AI?
Craft AI is an MLOps platform for data science teams.
To use Craft AI, the basic workflow with the platform is:
Choose a configured environment on the cloud provider of your choice
Create Machine Learning pipelines with your Python code
Deploy and execute the pipelines on environments running on Kubernetes
Monitor the performance of the models in production and the health of the infrastructure
The platform aims to be an end-to-end MLOps tool that brings together all the MLOps functionalities required for the successful implementation of an AI project. It is therefore composed of the following main features:
Environments
Machine Learning Pipelines
Deployments
1.2 Mission
Our mission is to democratize the use of trustworthy Artificial Intelligence on a day-to-day basis. How do we do this? By empowering Data Science teams to master their AI project from start to finish. Our vision of AI is that every project of Data Science should be in production, responsible and profitable!
To achieve this, we have developed a MLOps platform that allows anyone to put Python code into production, on a large scale, in a few clicks. We allow Data Scientist to deploy their models, choose their environments (development and production) and create pipelines to optimize their ML workflows.
How does it work? We will contain your code in step to allow you to create ready to production ML and DL pipelines in an environment adapted to each project’s needs.
1.3 History
Craft AI is a French company founded in 2015. We originally developed AI solutions for the energy, industry, healthcare, education and retail sectors. Our unique ability to deploy thousands of ML models at large scale with a focus on being always explainable, energy frugal and fair, drove us to develop a MLOps platform as a front end to our expertise. Today, with our MLOps platform, we can share our expertise in large-scale model deployment, at large scale.
1.4 What is MLOps?
Machine Learning Operations (MLOps), aims to provide an end-to-end development process to design, build and manage reproducible, testable, and evolvable ML-powered software.
Being an emerging field, MLOps is rapidly gaining momentum amongst Data Scientists, ML Engineers and AI enthusiasts. Following this trend, MLOps differentiates the ML models management from traditional software engineering like DevOps and suggests the following MLOps capabilities:
MLOps aims to unify the release/production cycle for ML and software application release.
MLOps enables automated testing of machine learning artifacts (e.g. data validation, ML model testing, and ML model integration testing)ML
MLOps enables the application of agile principles to machine learning projects.
MLOps enables supporting machine learning models and datasets to build these models as first-class citizens within CI/CD systems.
MLOps reduces technical debt across machine learning models.
MLOps must be a language-, framework-, platform-, and infrastructure-agnostic practice.
1.5 Benefits of using Craft AI
The main benefits of the platform for the users are:
No longer taking 6 months to deploy ML models in production but only a few clicks!
Allowing a Data Science team to be autonomous on AI in production without DevOps skills.
Enabling large scale production of Python code without refactoring to Java or C.
Automating the execution of the pipelines to save time for Data Science teams.
Ensuring an efficient use of computing resources and reduce the cloud bill.
Improving model performance over time by automatically triggering re-training pipelines when performance drops.
1.6 For which use cases?
Link to Tutoriels DS