KEYOU Designs Clean Hydrogen Engines on AWS for an Emission-Free Future

German automotive technology startup KEYOU is developing clean, sustainable vehicle engines using emission-free hydrogen fuel. But the company’s on-premises infrastructure was hampering its ability to innovate and deliver its technology quickly. It needed a more scalable and flexible solution that would provide more compute power in a cost-effective way.

To improve its ability to innovate and reduce costs, KEYOU’s engineers adopted a cloud-based workflow for computational fluid dynamics (CFD) simulations on Amazon Web Services (AWS). By running CONVERGE, a CFD simulation tool from software company and AWS Partner Convergent Science, KEYOU gained the ability to scale up to hundreds of processing cores when necessary and create higher-fidelity simulations for its engineers.

Since 2015, KEYOU has been redesigning traditional internal combustion engines to power emission-free commercial vehicles without compromising performance or range. The company aims to deliver its product and share its expertise with market-leading engine manufacturers. Distributing this technology is central to the startup’s mission: to use its years of engineering experience to help reduce carbon emissions from vehicles and pivot the automotive industry toward a sustainable future.

A key part of KEYOU’s development process is running CFD simulations of fuel combustion and hydrogen injection. In the past, KEYOU relied on local workstations to perform CFD simulations. However, its ability to do so was limited by the constraints of on-premises hardware. “I saw that we were running into bottlenecks with our workstations very quickly,” says Gidion Maniezki, engine simulation engineer at KEYOU. With a tight budget and big ambitions, the company looked to the cloud. To keep its focus on engine innovation instead of information technology infrastructure, KEYOU reached out to Convergent Science in January 2021 for its CONVERGE simulation software. That same month, the companies began working to develop a proof of concept on AWS.

KEYOU needed to implement its new solution quickly to avoid disruptions. The company’s CFD engineers consulted AWS solutions architects to build a web-based user interface for performing full CFD workflows. Relying on continuous support from AWS, KEYOU deployed a web front end using AWS ParallelCluster, an open-source cluster management tool that makes it simple to deploy and manage high performance computing clusters on AWS. While the processing cluster runs on AWS ParallelCluster, the company uses NICE EnginFrame—which simplifies the setup and running of technical and scientific applications—to power the graphical user interface.

“There is no comparison between the new interface and the old system,” says Jonathan Egerer, information technology operations manager at KEYOU. Usually, cloud deployments take 6–9 months to complete. Using AWS, KEYOU completed its migration in just 6 weeks. The company received $100,000 in credits through AWS Activate, a program offering free tools, resources, and training to startups to help them quickly get started on AWS. The credits helped KEYOU offset proof-of-concept costs while it built its simulation workflow in the cloud. AWS also provided an additional $5,000 in credits for AWS Business Support, which offers technical support and architectural guidance. “With the credits, we can enable the team members to develop, design, and perform at their best,” says Markus Schneider, chief operations officer of KEYOU.

Credits aren’t the only way that using AWS helps KEYOU cut costs. To meet its performance needs, deploying on-premises clusters would have been too expensive, but on AWS, KEYOU can use powerful compute servers with hundreds of processing cores only when it needs to and scale down at other times, resulting in significant cost savings. “We would never be able to run 600 cores cost effectively on premises,” says Schneider. On AWS, the company can be flexible. Instead of locking hundreds of thousands of dollars into physical infrastructure, this scalability makes the startup more liquid. “Using AWS, we can invest in the product rather than the infrastructure,” says Schneider.

Much of KEYOU’s compute workloads occur on Amazon Elastic Compute Cloud (Amazon EC2), a web service that provides secure, resizable compute capacity in the cloud. For simulation speed, KEYOU uses Amazon EC2 C5 Instances—specifically Amazon EC2 C5n Instances—which provide cost-effective high performance for compute-intensive workloads. To keep costs low, KEYOU also uses Amazon EC2 Spot Instances, which offer unused Amazon EC2 capacity in the cloud at a discount of up to 90 percent compared to Amazon EC2 On-Demand Instances. Previously, the most intensive simulations took up to 2 weeks to complete. On AWS, that time is cut to about 24 hours. Because running simulations on AWS is so fast, KEYOU can use one CONVERGE simulation solution license, further increasing savings. In the past, the company needed to run parallel simulations on multiple computers—each of which requires a license—to match the pace it can now achieve on AWS. “If we have compute power that’s 10 times faster, we can do 10 times as many simulations with a single license,” says Schneider.

KEYOU also used Elastic Fabric Adapter—a network interface for Amazon EC2 instances that enables customers to run high performance computing and machine learning applications—to enable better scaling. Throughout the process, the AWS team earned KEYOU’s trust by providing consistent guidance on system architecture, cost optimization, and running workloads on AWS. “When you have trust in your provider, you’re confident that it understands what you need,” says Schneider.

KEYOU has further ambitions for its automotive innovation on AWS. “We’re mainly delivering on-road solutions for heavy-duty vehicles, but we’re starting to look at off-road opportunities,” says Schneider. By using AWS, KEYOU gains the powerful cloud infrastructure it needs to work toward its mission of creating a sustainable future. “The combination of scalability, flexibility, and user friendliness makes AWS the best provider for us,” says Maniezki.