Dataiku

Seeing that one of their goals is to find ways to better answer consumer and business needs, the Global Engineering (GE) team was facing the challenge of improving processes and performance of the surfacing machines to significantly improve their production by using the increasing volume of data."We wanted a data science platform that would allow us to solve our business use cases very quickly. Thanks to Dataiku and its collaborative platform, which is agile and flexible, data science has become the norm and is now used more widely within our organization and around the world," said Cédric Sileo, Data Science Leader at Global Engineering, Essilor.

The Data and Analytics team at U.S. Venture began in 2018. At the time, the team was doing data warehousing and basic reporting, but soon realized they needed the right people and tools to do advanced analytics at scale — maintaining models and disparate data sources were going to become unmanageable quickly without them. That was their initial pain point: They had people that could have built solutions from scratch for DataOps (but no automation surrounding data collection, prep, and model connectivity) — but it wasn’t the time to reinvent the wheel.Further, the team’s data scientists and analysts were using a varied set of tools and coding mechanisms — one data scientist used R, one used Python, some analysts were using SQL while others used Python, and so on. Resultantly, the individual team members built their own components that lived in different places and were created via their own tools, saved on personal computers, with no visibility for other team members about where projects were and how they were created or functioned. This prevented them from supporting each other and collaborating on projects. 

PriceMoov’s challenge was that data originating from old SI systems, Oracle, or MySql was dirty and required a full-time developer to perform long ETL (extract-transform-load) steps in PHP for cleaning. Once cleaned, the datasets were painfully entered into a model, as they were custom-built pipelines. And once finished, the replication and deployment process for the next customer was taking weeks.