Data Scientists and the AI Revolution

Data scientists want to work faster and smarter everywhere: data access, data pre-processing, feature engineering, model training, and testing.

Watch Video

But data scientists also have a lot to gain from:

• Enabling a culture of operationalization: Pushing to go from prototypes to production is critical to success, and it starts with data scientists. A culture of operationalization is how data scientists will build trust – and prove – that data science brings real business value.

My greatest fear about data science…is that the penny will drop and people will realize that a lot of what has been promised is not yet possible.

Shaun McGirr
Head of Data Science and Business Intelligence, Cox Automotive

• Cross-team collaboration: When data scientists work directly with subject matter experts, projects are better aligned with business objectives and closer to generating value for an organization (read: support for data science initiatives – and data scientists themselves – will continue to grow). Plus, effectively incorporating business analysts in data projects can help shift time spent on data cleaning and preparation.

Watch Video

• Responsibility and white-box approaches: These subjects go far beyond ethics (though also critically important for data scientists) and touch on an approach to data science that is sustainable for years to come. For data scientists themselves, that means involving the right people, asking the right questions, and holding themselves responsible for practicing responsible AI.

Watch Video

• Automation: Automate data-preprocessing, feature engineering, and model training to quickly discover and build the machine learning and AI services needed to deliver value from data.
• Operationalization: Create a one-stop-shop in a visual interface to load models and serve requests in just a few clicks. Say goodbye to human bottlenecks and hello to quickly deployed models and user independence.

Watch Video

• Modeling: Save time with parallel model building and AutoML functionalities – see the performance of hundreds of models competing in real-time, and quickly identify the obvious winners (and losers).
• Reproducibility: In Dataiku, deployed code is reproducible – even unmaintained projects won’t fall through the cracks and fail.
• Flexibility and White Box AI: Create custom reusable components (plugins) out of existing code and leverage the tools (R, Python, Scala, etc.) and libraries (MLlib, H2O, XGboost, Scikit Learn, etc.) from the rich and ever-growing open-source ecosystem to build and run data services from scratch.

Watch Video