Customer Stories / Engineering, Construction & Real Estate


Facilitating Sustainable City Design Using Amazon SageMaker with Arup

Learn how Arup supported urban sustainability using Amazon SageMaker geospatial machine learning capabilities.

Built ML models

using geospatial data

Cut time to market

and cost for new projects

Off-loaded maintenance

of ML tooling

Improved recommendations

for sustainability

Reduced time significantly

in onboarding additional data scientists


Arup wanted to scale its data analytics solutions for clients in sustainable urban planning and construction—instead of building from scratch for each project. Arup’s clients expect rapid, efficient, accurate machine learning (ML) models that produce visually rich results. Arup wanted to build these models to better understand a broad range of relevant data and generate predictions to help clients make more informed, actionable urban-planning decisions.

Arup data scientists had built its Arup Data Platform using a suite of solutions available from Amazon Web Services (AWS). For data science and analytics, the company used Amazon SageMaker to complement its domain expertise and build, train, and deploy ML models for virtually any use case with fully managed infrastructure, tools, and workflows. To further enhance ML capabilities, Arup decided to incorporate a new offering from AWS, Amazon SageMaker geospatial capabilities. This enhancement provided specialized features within Amazon SageMaker that facilitated geospatial data analyses and the application of ML models on geospatial data. The enhanced Arup Data Platform streamlined and standardized internal processes, generated predictions from geospatial data, and improved Arup’s ability to help its clients make more sustainable decisions in urban design and planning.

construction and crane panoramic

Opportunity | Using Amazon SageMaker to Mitigate the Environmental Impact of Urban Projects

Headquartered in London and with 90 offices worldwide, Arup provides sustainable design, engineering, architecture, planning, and advisory services for its clients and partners. One of Arup’s priorities is to understand the factors that cause urban heat islands (UHIs), areas in cities that experience warmer temperatures compared to their rural surroundings. The company uses satellite imagery, Earth observation, and remote-sensing data to model and quantify the factors that influence UHIs. These help Arup to advise stakeholders on how to mitigate the effects of UHIs using environmentally conscious design and planning. “Sustainability is at the heart of everything that we do,” says Damien McCloud, Arup’s director and lead for geospatial and Earth observation.

Before they began using Amazon SageMaker, Arup domain specialists worked independently on each project, using open-source ML tooling that they customized and maintained. With significant jumps in the scale of data to analyze, projects faced increasing costs and complexity in their routes to market. Plus, teams were spending time reviewing open-source licensing, managing dependencies, and validating security standards of ML tool stacks on top of the data science work. They also had to continuously patch, upgrade, and maintain development environments manually for the duration of each project. Data scientists were keen to build a simpler solution on AWS to increase productivity and lower costs so that they could take on larger projects. “It’s a lot of work to build and maintain these tools,” says Mike Exon, Arup’s global cloud and data leader. “Across the firm, there were good opportunities to improve our approach and work more efficiently. We could do bigger projects and solve more challenges for our clients if we were better able to scale.”


Our use of Amazon SageMaker brings multiple layers of data together and asks complex questions at scale with volumes of data. We couldn’t do that before we turned to AWS.”

Damien McCloud
Director and Business Lead for Geospatial and Earth Observation, Arup

Solution | Standardizing ML Workflows Using Amazon SageMaker Geospatial Capabilities

Each year, Arup takes on more than 100 Earth-observation projects. The company wanted to optimize ML models that could isolate the impact of environmental factors, such as air quality, temperature, and building materials. Arup coordinates data collection and centralizes ML models through Amazon SageMaker, which builds algorithms and facilitates the production workloads so that Arup can deliver customer projects more quickly. The company based its Arup Data Platform data science capability on Amazon SageMaker, which relieved operational burden and facilitated collaboration among data scientists, whether they specialized in structured, computer vision, or geospatial data. “For Arup’s projects around UHIs, our use of Amazon SageMaker brings multiple layers of data together and asks complex questions at scale with volumes of data. We couldn’t do that before we turned to AWS,” McCloud says.

Arup wanted to use a broader collection of digital assets to generate more accurate predictions and create tangible recommendations for clients. It became one of the first users of Amazon SageMaker geospatial capabilities, using the new offerings for the preparation of data and analytics that involved complex datasets based on rasters, or Earth-observation images. Arup worked alongside AWS to run proofs of concept, with a goal to standardize workflows and create repeatable solutions. The Amazon SageMaker development teams incorporated feedback directly from Arup so that they could focus on features most important to customers as they optimized Amazon SageMaker geospatial capabilities. “The use of Amazon SageMaker and what we’ve done with the geospatial ML capabilities meets a sweet spot for us,” says Exon. “We now have a single, consistent platform across the data science community that supports the various specialists and simply lets Arup use combinations of data science specialisms appropriate to solve a client’s needs.”

The prebuilt, standardized set of tools—such as purpose-built Earth-observation operations, pretrained models, and preinstalled, open-source geospatial libraries inside Amazon SageMaker Notebooks, fully managed notebooks for exploring data and building ML models—alleviates work for the Arup Data Platform team, which otherwise would have had to compile, build, maintain, and secure complete tool sets for all specialisms of data science. “Amazon SageMaker geospatial capabilities gives us a ready-to-use, packaged solution, removing the need to build and integrate Earth-observation capability into the Arup Data Platform data science solution,” Exon says. “This saves us months of effort and delivers a richer solution in half the time.”

Arup frequently takes on new data scientists in the middle of projects, and the Arup Data Platform has reduced the time to onboard a new data scientist, a process that used to take weeks and now takes about 30 minutes. “Our route to market is simpler, and we have confidence that we know how much things are going to cost and how long they’re going to take,” McCloud says.

Outcome | Building Sustainability into City Planning

Arup plans to continue working to understand the nuances of UHIs and how the impact on cities and the environment can be mitigated through the use of data and advanced analysis tools. Arup also plans to expand the capabilities of the Arup Data Platform by incorporating additional AWS services. The company will be able to address an increasing number of use cases that benefit from specialized analytics, such as biodiversity management or sustainable land use. “Using Amazon SageMaker, we have delivered a rapid, rich technical capability,” Exon says. “But the real value we deliver through Arup’s data scientists is in support of sustainability for building modern cities. That’s the 40-year viewpoint of sustainability.”

About Arup

Arup is a global collective of designers, consultants, and experts dedicated to sustainable development and to using imagination, technology, and rigor to shape a better world.

AWS Services Used

Amazon SageMaker

Amazon SageMaker is built on Amazon’s two decades of experience developing real-world ML applications, including product recommendations, personalization, intelligent shopping, robotics, and voice-assisted devices.

Learn more »

Amazon SageMaker Notebooks

Amazon SageMaker offers two types of fully-managed one-click Jupyter Notebooks for data exploration and building ML models – SageMaker Studio Notebooks and Notebook Instances.

Learn more »

Get Started

Organizations of all sizes across all industries are transforming their businesses and delivering on their missions every day using AWS. Contact our experts and start your own AWS journey today.