Re­pro­ducible data science pipelines in produc­tion a workshop with Tania Allard

How many times have you developed a model or a data application and tested it locally just to find out that it breaks in production? How many times have you seen a data science project that is ready to production but the closer it gets the harder it is to answer the following: “Why did the model predict this?”

These are two common issues faced by many many data scientist across the globe.

On one hand, explainability is essential for the trust of a model and its predictions. It also helps prevent situations in which nobody understands why a prediction is made. On the other hand, as more and more data-intensive applications are created, there is also a higher need for practical data operations and processes to improve the deployment of machine learning models.

In this workshop, you’ll learn how to level up your data science workflows with some practical DevOps! We will focus on how to improve the reliability and quality of your data applications preparing them better for production deployment and consumption.

We will build an end to end machine learning pipeline focusing on automated testing, integration and delivery, increasing its reliability.

What are the main takeaways? You will acquire an understanding of DataOps and how these can improve your data science workflows. We will focus on model explainability without compromising its accuracy. As we move over to the examples, you will better identify the many challenges faced during the production implementation of data applications and how these can be mitigated through best Ops practices. By the end of the workshop, you will have the knowledge required to automate the delivery of their data products, increasing their productivity and the quality of their work.