Customer Case Studies: Enabling Code Portability and Simplifying Integration of AI on MCUs

As embedded devices become more sophisticated, companies face increasing challenges in software (SW) development, integration, and deployment. The traditional approaches involving monolithic firmware or embedded Linux are resistant to innovation as they require significant engineering effort to support multiple hardware (HW) platforms, are difficult to program, and 

challenging to update and maintain.  

Atym is transforming how companies develop and manage software for resource-constrained embedded devices with our edge orchestration solution that enables lightweight containers based on WebAssembly to provide benefits including modularity, enhanced flexibility, accelerated 

development, and security. Atym makes it possible to run containerized software on MCUs with as little as 256KB of memory to reduce engineering friction, increase security, and simplify ecosystem collaboration while protecting IP. It can also serve as an alternative to Docker for constrained MPU- and MPU/CPU-based devices in order to free up memory for applications or reduce product BOM cost.   

Customers leverage the Atym Hub and associated tooling to build, deploy, manage, and secure containers that are deployed on the Atym Runtime on their devices. The Atym Runtime is based on the open source Ocre project in the Linux Foundation which affords transparency, eliminates lock-in and helps scale device support. 

The Atym solution 

Our solution is suited for use cases spanning all industry verticals and the consumer space, but as I will be attending Hannover Messe next week, I figured I would share some of our recent engagements with global Fortune 500 industrial OEM customers. In this blog, I will explore a few use cases where Atym has empowered customers to overcome their embedded development challenges with MCU-based devices in areas of software portability and integrating AI models. 

Find us at Hannover Messe 2025!

Since making a bet on WebAssembly a few years ago as our core enabler of our solution, we have seen a clear pattern of organizations of all types experimenting with the technology to address the historical challenges with embedded development. This is no exception in the industrial space. This recent paper from ARC Advisory Group walks through the benefits of WebAssembly for modernizing embedded industrial software and features Atym.  Also linked from that page is a related joint webinar that provides great background on the opportunity. 

Enabling Software Portability and Stability Through Fractional Updates

One of the biggest hurdles in embedded software development is the lack of code portability. Traditional embedded development tightly couples code to specific HW, often taking months to migrate existing code across different products and HW or to support a multi-source silicon strategy. This results in increased development costs, longer time to market, difficulty in maintaining code over time, and potential supply chain issues like we saw during the height of the Covid pandemic. A major aspect that customers appreciate about our solution is how it helps them completely revolutionize the way they approach embedded development to address these challenges.  

Atym containerization vs. traditional monolithic firmware 

We recently helped a global industrial measurement leader increase the portability of their code to truly realize the promise of write once, deploy anywhere with their embedded device offerings. This customer has a centralized engineering group that develops SW functions (e.g. connectivity, 

application-level protocol support, security elements) that are leveraged by their different global business units for their various products.

Each of the central engineering team's implementation of a function requires writing code for the different MCUs and MPUs/CPUs used across the company’s expansive portfolio. The bespoke development of products across their catalog greatly complicates the group’s development process and ability to maintain code for a complex matrix of permutations.  

Historically, when a business unit (LOB) within the organization leverages one of the core engineering group’s SW functions, they have needed to integrate the source code into their device’s monolithic firmware so it can compile and run correctly. When the core engineering group finds a bug or security issue, the updated code requires all the business units to recompile, retest, and then update the fielded devices. Many of the company’s end customers adhere to strict compliance requirements, so the required retesting is extensive and costly before the entire device SW can be updated in the field. In addition, many of their customers cannot experience any downtime to reboot a device for a SW update, so many plan a SW update with scheduled maintenance which could be months to a year out. 

With Atym’s WebAssembly-based container solution, the core engineering group now only needs to develop one version of each SW function since it is abstracted from the underlying chipsets by the Atym Runtime. Our solution enables them to compile their code as a containerized Wasm binary that is portable across chipsets, eliminating the need for redundant development to support all of the different HW platforms used by their internal customers in the LOBs. 

Moving to Atym will save the core engineering group thousands of hours annually in undifferentiated development to support multiple chipsets with the same functionality. 

The Atym Runtime also enables each of the business units to move away from monolithic firmware and the headaches of figuring out how to integrate the core engineering group’s SW functions into their device code. They can now implement their devices’ SW functions and applications in Wasm containers which are chipset agnostic. When an application is deployed to a device, the Wasm 

container binary is compiled ahead of time on the Atym Hub to the native instruction set of the specific MCU or MPU/CPU on the targeted device. This has the potential to eliminate 100’s of hours of project delays per year for each business unit, allowing their teams to focus on innovation instead of integration hurdles. 

Atym containerization and hardware abstraction enables code portability and reduces churn

In addition, because our solution enables fractional updates of the containerized apps, each business unit now only needs to focus on the specific applications that require updates instead of recompiling and retesting the entire code base for the device. Fractional updates of a specific container also do not require rebooting the device, which is critical for many of their end 

customers to avoid downtime just to schedule a SW update. This enables their customers to update devices faster with highly focused code changes that de-risk concerns with respect to compliance regulations and costly downtime.  

Simplifying Cross Org Collaboration to Integrate On-device Machine Learning

The addition of machine learning (ML) to edge and IoT applications is becoming increasingly important to drive real-time response, conserve valuable network bandwidth, increase security, and preserve privacy. The rise of edge AI/ML is causing a necessary change in the industry's approach to embedded development because the models will need to be updated over time as they are retrained. This is a fundamental difference from a traditional embedded device like a wireless sensor that is fixed in function and may only be updated in the field if absolutely necessary for a security patch. 

However, integrating ML models into monolithic firmware is a complex and time-consuming process. Many ML models are developed in Python, while most embedded systems use C/C++. This creates a significant integration challenge. Even with ML Ops tools that generate C code, the challenge still remains for ML engineers, who are not familiar with C/C++ or embedded HW, to figure out how to integrate their code into the targeted device’s monolithic firmware. 

Our customers appreciate how the Atym solution greatly simplifies ML model integration for their products.  As an example, we have been working with a leading industrial automation OEM that is adding AI capability to one of their MCU-based sensors to analyze collected telemetry data. They have a core embedded development team familiar with programming firmware in C, but their ML engineers in a different group are familiar with developing models in Python and leveraging 

modern CI/CD practices with container technologies like Docker. The ML engineers do not have experience with embedded HW, real time operating systems, or C programming. 

These two teams ran into considerable project delays when trying to integrate a Python-based ML model into the monolithic device firmware. In addition, both teams knew the code integration challenges were going to become even more complicated in production because inevitable updates to the ML model as it was re-trained would require the entire device firmware to be recompiled, retested, and redeployed on devices in the field.    

With the Atym solution, our customer is able to move away from monolithic firmware written and complied entirely in C to a modular approach utilizing containerized applications deployed on top of the Atym Runtime. The ML models can now remain based in Python and run alongside existing device code written in C, in addition to future apps programmed in other language choices (e.g. Rust, Go) as needed. This lowers their development costs while accelerating time to market, and simplifies adding new capabilities over time. With the HW abstraction provided by the Atym Runtime, their ML engineers no longer have to worry about understanding the embedded HW intricacies to get their models to work optimally. In fact, they do not even need to change their workflow - they can simply work with Atym in the same way that they would leverage Docker.

Teams can focus on their core competencies and apps such as ML models can be fractionally updated without downtime 

As with the first example, the ability to fractionally update containerized ML models eliminates the need to reintegrate code, recompile the entire code base, and conduct complete device retesting. This shift drastically accelerates their CI/CD pipeline, reduces the update cycles from months to days, and allows them to release valuable updates to their customers more quickly.    

Conclusion 

Atym empowers businesses to accelerate innovation with resource-constrained embedded devices while reducing cost and complexity. Companies spanning OEMs to end users are evaluating and adopting our solution to turn firmware into modular, containerized software and are seeing significant gains in efficiency, flexibility, and maintainability.  In turn, they like the insurance policy afforded by our open core business model. 

Atym benefits MCUs and resource-constrained MPU/CPUs in myriad use cases 

These are just a few customer case studies that illustrate the transformative impact of our solution 

in the areas of increasing software portability and simplifying ML integration. In future blogs, I will 

talk about other key customer benefits we are seeing with our solution, such as simplifying the implementation of a robust cybersecurity model.

Let us know if you are interested in learning more about how we can help you modernize your embedded software development practices to build more agile, flexible, and secure devices and 

systems. Please also drop me a line if you happen to be at Hannover Messe next week – I would 

love to connect!