Want to Stop Rebuilding "Expensive" Parts of your Flow? Explicit Builds are the Answer!READ MORE

Eulidia - Enabling a Leading Luxury Group and Its Brands to Gain AI Maturity at Scale

Name: Arnaud Canu

Title: CTO

Country: France

Organization: Eulidia

An exciting journey from Data to Business Performance! Eulidia relies on both technical expertise on AI and Cloud Architecture, and performance management to turn your data into innovation.

 

Awards Categories:

  • Value at Scale
  • Most Impactful Transformation Story
  • Partner Acceleration

 

Business Challenge:

Our job at Eulidia is to help companies become efficient and innovate with data. A few years ago, when we first started working with a leading luxury retail group, we quickly realized their main pain point was the lack of any AI platform - use cases were developed in Python on local machines, industrialization meant fully re-implementing algorithms, collaboration between data scientists was extremely difficult.

The group encompasses  multiple luxury brands, each with very different skills and AI maturity. The central organization aims to help those brands quickly gain AI maturity and deploy use cases. It quickly became obvious that reaching a satisfying maturity level in each brand would require a unique AI platform, easy to deploy and apprehend by users, allowing the mutualization of  use cases and normalized processes.

My role at that time was to lead our consulting team and help design this future AI platform. By benchmarking existing platforms and comparing their capabilities to the client needs, we quickly showed how Dataiku would be the best solution to deploy.

The retail organization started with a very simple Dataiku deployment, on which my team and I re-implemented one of their existing AI use cases – this deployment quickly revealed huge gains in terms of development time, reliability, maintainability, and collaboration possibilities. Based on this first success story, a Dataiku rollout was decided as the main AI platform.

Business Solution:

After that first deployment, we mainly worked with the IT team to design the target organization for their data team and the associated processes, from first developments to industrialization and deployment in production. These processes were implemented on the first few use cases, which allowed us to train their developers through the implementation phases. This second success story was concluded by the formalization of a “Dataiku Framework” defining the governance, naming conventions, underlying technical environment and how to efficiently use it, best practices, etc.

Based on this framework, we mobilized an Eulidia implementation team of 13 people, and we started actively implementing use cases. Together with the IT team, we deployed and configured 24 Dataiku nodes, on which we developed data pipelines, AI algorithms, and dedicated plugins. Based on these developments, we helped the client design a set of plugins for platform administration and monitoring, as well as dedicated tools for Brands Business Analysts.

Use cases rollout quickly motivated business users to join this AI initiative. In order to help each brand gain AI maturity, we designed a dedicated training program, which my team delivered to 110 business users and developers, across 12  brands. Based on this third and last success story, we are now helping the client on their continuous improvement of the platform, delivering new capabilities to brands such as Kubernetes clusters, CI/CD pipelines, etc. In parallel, we now work with business teams to help identify and implement high value use cases, using the available Dataiku platform.

Business Area Enhanced: Marketing/Sales/Customer Relationship Management

Use Case Stage: In Production

 

Value Generated:

One immediate gain of our work is the adoption by each business team. Data and AI used to be difficult to implement and deploy, it’s now easily available for business analysts and data scientists. Plugins and frameworks help quickly build new use cases, while dedicated data warehouses and cloud infrastructure reduce experimentation costs.

The configuration process and tools we developed allows for easy, quick and automated nodes deployment for new brands. Standardization allows for projects and models mutualization, which helps even small brands reach high maturity usages!

The governance we defined and deployed helps multiple populations efficiently collaborate, from business-oriented teams to more technical experts. Controls and best practices are enforced and ensure smooth delivery to production. Thanks to that approach, multiple use cases have been deployed into production since the beginning of the year, at a very quick pace.

 

Value Brought by Dataiku:

The Dataiku platform offers multiple advantages for such a large deployment. Among those, we especially benefited from the following aspects:

  •  Collaborative approach,
  • Cloud integration,
  • Mutualization and deployment speed,
  • Governance capabilities.

First is the seamless experience for collaboration: users from the retail group have very different skill sets and expectations. Using Dataiku, highly technical developers can collaborate with business-oriented users, and the resulting projects can easily be deployed and monitored in production by dedicated teams.

Another huge advantage was the integration capacities with the client cloud environment. Our client uses the GCP cloud for both data storage (data warehouses, data lakes, etc.) and computational capacities (such as Kubernetes, virtual machines, or SQL databases). By design, Dataiku nodes can be deployed on top of such infrastructure. Such an approach allows for quick and painless integration.

One main objective was to mutualize as many assets as possible, and quickly deploy use cases into production. The Dataiku extensibility (for example through plugins) made such mutualization very natural, which the embedded delivery tools (deployer, Git integration, etc.) allowed us to integrate with the client CI/CD tools and pipelines.

Finally, defining and enforcing models and project governance allowed us to combine business needs with IT and production expectations. We were able to easily define user profiles and responsibilities, then to enforce delivery processes such as code reviews, documentation, etc.

Value Type:

  • Improve customer/employee satisfaction
  • Increase revenue
  • Save time
Share:
Version history
Last update:
2 weeks ago
Updated by:
Contributors