The Machine Learning Platform of Stolt-Nielsen

Stolt-Nielsen, a leading player in logistics and shipping, deploys AI technology to increase efficiency and improve service delivery. Predictive Machine Learning models were developed both internally and externally, but this method presented challenges in maintaining standards, scalability, and manageability. As a solution, Motion10, an HSO company, built a Machine Learning Platform in Azure that connects to Stolt-Nielsen's previously developed data platform. In this case study, discover how using a centralized platform brings strategic benefits to the business.

Pioneer in logistics and shipping

A pioneering spirit and a knack for anticipating new opportunities courses through Stolt-Nielsen’s veins. The organization is the world's largest operator of advanced chemical tankers, a global provider of secure bulk liquid storage services, and provides door-to-door transportation services for bulk liquid chemicals and food-grade products. Furthermore, Stolt-Nielsen offers innovative services in high-tech and sustainable fish production. Throughout its sixty-plus years of existence, the organization has developed a deep understanding of the operational challenges and opportunities that technology can provide for innovative solutions in the logistics and shipping industry. Stolt-Nielsen ensured that groundbreaking technologies and concepts were developed, improving efficiency and safety in the industry. As an example, Stolt-Nielsen was the first company to develop and implement integrated tankers with separate tank systems to reduce the risk of leakage and contamination.

Standing out with technology

Today, Stolt-Nielsen's distinguishing factors include minimizing risks, ensuring the highest standards of safety – both for their employees and the environment – and providing plenty of room for innovation in IT. Erik Visser, Director of Architecture & Engineering at Stolt-Nielsen: "Our ambition is to be at the forefront of our industry by deploying innovative technology such as data applications and AI (Artificial Intelligence), to help us use our resources more efficiently and serve customers better. We have clear goals in mind and considerable investment is being made in this area, across the organization.”

Erik: "We embarked on a journey with the team to develop the first integration on the data platform. This step was aimed at expanding Business Intelligence capabilities and more traditional workloads. To make this happen, a cloud-native data warehouse was set up in the Azure Cloud, which is part of a solid foundation that we continue to build on.”

Although more and more value is being created with data – partially through the training of non-IT departments in the use of Power BI – Stolt-Nielsen aspires to elevate its technological capabilities in 2022, to tap into the potential of advanced analytics.

From analyzing to predicting

The ability to forecast supply and demand, pricing, cargo capacity and optimal routing for ships and tank containers represents a major strategic advantage for Stolt-Nielsen over their competitors. The ability to predict the right heating temperature – based on weather patterns – to heat tanks in Stoltharbor’s terminals (one of the divisions in the organization) could yield substantial cost savings. Moreover, it contributes to sustainability goals, which is a key driver for Stolt-Nielsen. But what does it take to shift from looking back to looking forward with the help of innovative technology?

Erik Visser reasons: "Because of the setup with various external partners, we were unable to develop AI-applications in a responsible and scalable manner within our own environment. In order to gain more experience and skills in this area over time, we wanted to centralize the process to our own environment.”

Motion10, an HSO company, delivers an architecture and framework for the Machine Learning Platform in Azure and adapts some of Stolt-Nielsen's external models, so they can be built on the new, centralized platform. Next, the new Machine Learning Platform with its built-up models is securely connected to Stolt-Nielsen's existing data platform.

The current structure forms a new foundation on which new solutions can be built with scalable computing power, insight into models (for the purpose of Responsible AI), and easy deployment. As a result, Data Scientists have the right tools at hand to be even more productive in their work.

Implemented models with reliable outcomes

While many organizations get stuck in the experimental phase of training models, Stolt-Nielsen have actually implemented them, and run several models in production. "It is important to have confidence in what a model suggests" Erik explains. "If it is reasonable and explainable, it contributes to adoption and usability. The ability to test and manage models has improved, largely also due to the Machine Learning Platform.”