“I don’t need to know AI; I’m an embedded systems engineer!”
This statement may have made sense a few years ago. However, today, if you still hold this attitude, you will miss out on a key technology that is fundamentally changing the way software and system design is approached, and its influence will continue to dominate for decades to come.
AI is no longer the exclusive domain of cloud engineers and data scientists. While AI-on-chip remains out of reach for many microcontroller (MCU) level systems, for embedded developers, ignoring AI is equivalent to missing out on a powerful tool that can completely transform the design, development, and delivery of firmware.
No matter if you focus on real-time control loops, safety-critical firmware, or low-power Internet of Things (IoT) devices, AI is no longer an optional feature. It has become a key driver for faster and smarter firmware.
This article will introduce three AI tools that I believe every embedded developer and manager should learn about and explore.
Ollama: Run LLM Locally
Ollama is an open-source platform that allows you to run various large language models (LLMs) locally on your own hardware, such as LLaMA, Mistral, Gemma, etc., without relying on cloud services, eliminating privacy risks, and fully leveraging the power of AI locally. (Did I mention? It also doesn’t require a subscription fee.)
Why design it this way? The benefits are clear. First, Ollama emphasizes security by design. Since it operates entirely in an offline environment, it ensures that your source code, files, processes, and any proprietary knowledge will not leak. (Of course, this doesn’t mean you can use insecure scripts without issues.)
Secondly, Ollama is open-source, meaning it is highly flexible and free. You can search for suitable models based on your needs and download them directly to your machine. This is not the only way; you can also find open-source models from platforms like Hugging Face and pull them into your system.
Finally, you can customize and fine-tune the models. You can train and adjust open-source models to automatically generate peripheral drivers, establish testing frameworks, or even write technical documents in your unique style.
If you value security and want to continuously use LLMs in an offline environment, then Ollama is your best choice. One of my favorite aspects is that it allows you to freely experiment with various language models without being restricted by specific vendors, which offers much more freedom than many online platforms.
Of course, there are a few points to note. First, it does not have real-time customer support. When you encounter issues, you can only rely on your device and community assistance. Secondly, to achieve optimal performance, you need to have a local GPU or server.
If you want to introduce AI without sacrificing security and control, then Ollama is an ideal starting point that can seamlessly integrate into your workflow for developing faster and smarter firmware.
Agentized IDE: AI-Assisted Integrated Development Environment
The so-called Agentized IDE refers to the next-generation development platform integrated with AI agent capabilities. These IDEs are mostly based on Visual Studio Code and are enhanced with various plugins to expand AI agent capabilities, significantly improving efficiency for developers in their daily work.
Currently, there are several IDEs on the market with AI integration features, such as Cursor, WindSurf, and Fire Studio. These development environments come with built-in AI assistants that can understand your program code and development environment.
So, what kind of functionalities can these AI agent IDEs actually provide? Here are some core capabilities:
-
Intelligent code completion and refactoring suggestions
-
Real-time suggestions based on project history
-
Automatically generated test code, files, and frameworks
-
Script building and CI/CD configuration suggestion support
-
Interactive debugging support and search functionality
These features are quite practical for all embedded developers. Embedded development often involves repetitive and detail-oriented low-level work that does not necessarily require your full attention. With Agentized IDE, you can delegate these tedious tasks to AI and focus on system-level design decisions.
You might wonder, “My work is deeply embedded; can an AI assistant really help?” In fact, even in such deep development work, AI IDEs can provide significant assistance in creating scripts and developing tools (such as one-off IDEs or analysis tools).
Of course, these intelligent IDEs will not replace your compiler or debugger, but they can significantly accelerate your overall workflow, allowing you to complete more valuable designs and validations faster.
Codex: LLM Designed for Programming Languages
Codex is an LLM developed by OpenAI that focuses on programming language processing, capable of understanding and generating code in over dozens of programming languages, including C, C++, and Rust, which are common in embedded systems.
Codex is trained to automatically complete functions, translate logic across languages, and even explain compilation error messages.
For embedded developers, Codex can provide assistance in multiple areas. First, it can be used to write low-level drivers that interact with hardware. For example, you can ask it to generate drivers for SPI, I2C, USART, etc. (Yes, many teams rely on code provided by manufacturers, but if you need further optimization, Codex can help you refactor a better-performing version.)
Secondly, during development, it is often necessary to create prototypes of application modules in different languages to validate functionality. For instance, you might first create a filter in Python and then translate it into C or C++ for deployment. Codex excels in language translation, making it very suitable for this purpose.
Finally, in terms of our embedded applications, Codex can also assist you in generating unit test templates or even directly creating test cases.
Of course, you may have concerns: “Is it risky to let AI write my code? Doesn’t that mean handing my job over to machines?” In reality, Codex is not a threat but a tool. Like an assistant, its value depends on how you guide it. For embedded work, it will not replace your design thinking or timing analysis skills, but it can significantly accelerate those daily tasks.
Codex is like an ever-tireless intelligent coding assistant that can help you build stable, production-ready high-quality firmware faster.
Conclusion: The Role of AI in Embedded Development
For embedded developers, it is easy to overlook AI. After all, AI may not run directly on the target systems you design—but it can and should be integrated into your development process.
AI can provide a powerful and easily accessible set of tools that can fundamentally change the way firmware is written.
The tangible benefits these AI tools can bring include:
-
Automatically generating program code for daily tasks, such as peripheral drivers, unit tests, and technical documents
-
Using AI for error explanation, issue tracking, and repair suggestions to accelerate the debugging process
-
More accurate code reviews that intercept errors before peer reviews
The purpose of these tools is not to replace engineers but to free us from tedious, repetitive, and guesswork tasks. Because in the modern embedded systems field, only those developers who can build firmware faster and smarter will be the winners.
END
Source:EDN Electronic Technology Design
Copyright belongs to the original author. If there is any infringement, please contact for deletion..▍Recommended ReadingHow much does a legitimate Keil cost?Microcontrollers can output PWM like thisSharing an embedded development debugging tool!→Follow for more updates←