OHRA: Building for the Future of AI With Speed & Control

Prior State: “Manual, Time Consuming, Difficult”

Before implementing Dataiku, it was manual, time consuming, and difficult for the central OHRA data team to deploy models in collaboration with other teams (e.g., the actuarial team) and hand them off to IT for deployment. 

At the same time, the data team was faced with increased demand and enthusiasm for data-driven solutions. They therefore sought a solution that would accelerate their ability to execute on data projects for the business while also ensuring the solutions met Responsible AI standards both for their customers and for regulators. 

Enter: Dataiku

The central data team at OHRA used Dataiku to develop a robust, integrated ecosystem that exponentially speeds up the company’s ability to execute on data projects that bring value. 

Some of the highlights of OHRA’s solution, powered by Dataiku, include:

• Standardized data and model components within feature stores, which accelerate time-to-value by creating a common base for all teams to have consistent data naming, standards, and definitions.
• Reusable plugins, which generalizes the code that the data team writes so it can be reused more easily, empowering team members to collaborate on projects regardless of skill level. The central data team at OHRA has also developed macros to standardize templates in accordance with its governance framework.
• An operationalized governance framework to maintain trust and control over data and data products. Thanks to Dataiku governance features such as prescribed controls, checks, processes, and roles and responsibilities, data scientists are very clear on what is expected from them at every point in the development process.
• Bias assessment applied to every use case that impacts customers (without additional time spent manually writing code or running queries). For example, the Model Fairness Report plugin in Dataiku is an integral part of OHRA’s data governance, making assessments reproducible and comparable between projects.
• Easy deployment thanks to Dataiku APIs — the data team is now in control of deploying their own models without having to handover code or model artifacts to the IT team, saving time in discussions and testing. The IT team simply calls the API through a mutually agreed upon specification. If the model needs to be updated or changed later on, the IT team can simply keep calling the same API without much additional effort. The controls built into Dataiku, together with the embedded governance, provides the IT organization with trust that the data product they are integrating is built responsibly and reliably.

These features bring faster time-to-value for all use cases. For example, automations introduced in the claims process make claims handling more efficient and help customers get answers more quickly. At the same time, with governance standardized and embedded in the development process, OHRA can ensure their customers are treated responsibly and fairly, no matter what the business use case for machine learning and AI.

Building for the Future of European AI Regulations

Without the move to Dataiku as an integrated platform solution, the data team at OHRA would not have been able to meet the growing demand for data products from the business. They are now able to adopt a “fail fast” strategy, quickly testing and prototyping to arrive at efficient business solutions. Everything — from model design to deployment — is faster. 

For example, development time — as input data is now largely centralized and standardized, there is no longer a need to do input data reviews to ensure data joins were conducted correctly. 

In addition, using visual, reusable components (either through standard Dataiku functionality or custom coded plugins) makes iterative development much faster. Not having to look up and write out code speeds up the process, plus it makes code checking much easier because the code used has already been approved. 

Dataiku makes developing, checking, and bringing data products to production much more fun. Using the visual interface and recipes eliminates a lot of tedious code work and gives the team more time to spend on what’s actually important: helping our business and customers.

– Antal Nusselder, Lead Data Scientist, OHRA

Deployed models are also much more scalable — using APIs instead of batch processes provides answers or predictions where and when they are needed. Using Dataiku, the underlying architecture is also easily scalable when demand increases.

Now that OHRA has moved from manual scripts to an integrated pipeline solution with enhanced collaboration, models that would have taken one year to deploy in the past now can take as little as two months. This reduction in effort to deploy comes from:

1. Reduced time required to align priorities across different teams to facilitate handoffs. 
2. With the APIs set up within Dataiku, OHRA has a better link between the AI teams and IT, reducing the need for retrofitting and testing when moving toward deployment. 
3. Removing the need to build new APIs from scratch for every model built.

But of course with the impending EU AI Act, it’s not just about scalability — it’s also about minimizing risk. With the implementation of custom metrics, the use of the Dataiku Model Fairness Report plugin, and the implementation of an AI Governance framework within Dataiku, OHRA is able to apply stricter, standardized, and more transparent governance with little additional time spent.