Introduction

CodaLab is a powerful open source framework for running competitions that involve result or code submission.  Bring to your team a tool to practice their Data Science skills without leaving your organisation. You can either participate in an existing competition or host a new competition.

Duration: 1h00
Format: présentation

Presentation

The objective for EDF R&D is to be able to set up in-house competitions on topics such as deep learning, machine learning and data science in general. Codalab thus makes it possible by proposing to launch scripts via a web interface so that the scripts will be launch on machines in an automated way and to be able to make comparisons of performances between the different scripts.

Organizing competitions internally will revitalize group life, motivate team members even more by doing competitions in team. It would also help develop a good competitive spirit and push, for exemple, data scientists to write more and more powerful codes.

At the beginning I will talk about the architecture present in EDF R&D, then explain in more detail the circumstances that led to the emergence of this topic. Then I will introduce you to the technology, its advantages and disadvantages. I will explain the procedure to follow to successfully install codalab server. I will also present the difficulties we face during the installation.

Author

Robert Walid SOARES is a junior Data Engineer and Data Scientist. He just graduated, doing his last year in apprenticeship at EDF as a Data Engineer, and joins the Adaltas team in 2018. He administrated the HDP clusters and also worked on POCs (Proof Of Concept) with Apache NiFi and Apache Atlas.