Empowering at the Edge: the "Arduino way" to AI

What if AI felt like a door you could open, not a wall you had to climb? We dig into how Arduino’s approach—accessibility first, power when you need it—turns the edge AI buzz into a concrete path you can follow, whether you’re a student with a starter kit or an engineer shipping to a fleet.

We walk through a practical four-step journey: try AI through no-code experiments, understand it with pre-trained models, train by fine-tuning or starting from scratch with your data, and build something real that lives beyond a demo. Along the way, we unpack a core principle we call “abstraction without obfuscation”—removing friction while keeping the logic transparent—so you can inspect, modify, and truly own the systems you create. That design philosophy shapes everything from our open hardware portfolio (TinyML-friendly MCUs up to Linux-capable MPUs) to our integrations with popular AI frameworks and community-driven libraries.

You’ll also hear how cloud-native developer tools streamline the messy middle: browser-based workflows, single-device to fleet deployments, secure OTA updates, data collection for predictive insights, and closed-loop model improvement. Plus, we introduce our AI assistant as a coach that explains code, diagnoses bugs, and helps optimize for memory and speed—turning dead ends into learning moments. Real-world validation from a 35-million-strong community and enterprise teams, including automotive innovators, shows how openness and cohesion accelerate the leap from idea to production.

If you care about AI that empowers rather than intimidates, this conversation lays out the playbook. Subscribe, share with a teammate who loves to build, and leave a review telling us the project you’re dreaming about—we might feature it next.

Transcript

Accessibility Over Power

SPEAKER_00 0:06

Good morning, everybody. How's it going today? It's so great to see such a great audience and so many people here today. So, um, a couple of years ago, uh, when I was entrusted uh with leading Arduino Innovation, I had a long shot uh with one of the Arduino co-founders, and uh he told me something that uh I still think about uh almost every day. What he told me is that uh um Arduino DNA uh has never been and will never be about uh you know making uh you know the most powerful technology, making the most powerful hardware, making the most uh you know powerful tools. But it was all about uh making uh technology radically accessible to everybody and to anybody in the world. Um and when it comes to edge AI and when it comes to AI in general, um one of uh you know the main challenges that uh that we are seeing as uh you know the pace of innovation in this technology is so fast uh is that we get people risk of being left behind. Um we believe at Arduino that uh technology and especially AI should empower people, not intimidate people. Uh but too often what we are seeing is that uh uh you know AI is locked down layers and layers of complexity that are brought uh right uh in front of developers uh with the idea of making the latest and greatest available to everybody. Um at Arduino, though, is uh and you know, has been from the very beginning. Arduino is uh uh has turned 20 today, uh this year. And uh our mission from the very beginning has always been about uh uh enabling everybody to enhance their lives uh you know by giving them access to digital technologies. And we believe, and I believe that uh you know, in the same way that for the last 20 years uh we have been on a journey to try and uh uh you know really lower the access bar to electronics, uh to coding and uh you know to you know education in STEM. Uh the challenge uh for the next uh you know five, ten, fifty years uh is uh you know to make uh AI usable, understandable, uh, you know, to make it valuable to anybody, whether they are in the classroom, they are in the factory floor, they are in a lab, they are in space, they are uh you know just using at home, using some product. Um the when we come to you know the point is how we are solving this uh um the the the idea is uh you know to rely really on a bunch of pillars, right? Arduino is a platform that is built on simplicity is on simplicity and on some uh you know really core pillar. Uh in the last 20 years, uh we built an amazing community, and today we count uh you know almost 35 million active developers across the world, and we are adding uh you know more than one million developers every year, especially in emerging countries. Um 1.5 million of these developers are actually students uh that every year through curricula are you know getting their first taste uh of you know the newer technologies. And of course, uh these students uh are more and more already used to you know having a smartphone, uh, having devices at home, right? And so they are much more uh uh you know uh you know the requirements that they have, right? And uh you know the need uh that you know is there to make them uh engaged, uh, you know, the bar is higher and dire for us. But still, we want them to understand the technology, not just to you to be mere users of uh of it. And then enterprises, right? We have over 2,000 enterprises today that are actively deploying uh you know either on their factory floor or in their product portfolio, Arduino technology as part of their product and their value proposition. So from hobbyists to industrial applications, um, you know, we really uh you know focus on building a set of tools and uh uh tool chains uh uh and cloud uh and learning content with the goal of uh you know creating one connected experience across uh you know across the journey. Uh what we built is you know what I like to think as uh uh you know modular but at the same time cohesive uh you know stack, you know, something that encompasses and sponsors from hardware to firmware, uh, from software to cloud, um, but includes uh learning experiences and content and places where people can go, ask questions, share their knowledge, and be able to engage with each other. Um you know, it's you know a set of tools that people can use uh to prototype, uh, but at the same time is the same set of tools that you know people can use uh you know to scale and go in production uh you know with it. The result that we are aiming for is uh you know to enable people to go to be on a frictionless journey from the idea to the result of that idea, to really appreciating, benefiting from the result of that. Um there is a concept uh that I think that explains that that I like to call abstraction without obfuscation. Uh, this is a core principle that I try to put in everything that uh we are doing and are doing. The idea here is uh abstracting the complexity of the tools, but don't not hiding the logic behind those tools. Um key element is that we want developers uh to always feel in control of what they are doing. Whether they are beginners, and so they get uh you know access to the you know the easier experience, but also when they get uh you know more you know expert and to want and they want to get deep into what they do. They want to kind of you know open up uh uh you know the the hood and see exactly how the sausage is uh you know is made. Um what we want to do here is to build systems that are intuitive on one side, but uh something that people can open, can understand, and can even modify as deep as they want if they choose to do that, if they want to do that. And this is why everything on almost everything that we do is released also as an open source project, right? Uh not only for you know enabling people to understand what's inside, but also to empower them to be able to you know take it from there, take full control to the point of being able to completely modify everything from the ground up. Um I talked about journey. When it comes to AI, to edge AI, the goal that we have is really to set people on a journey, right? And uh, you know, the journey of a developer through AI, uh, we believe is uh you know based on four different steps. First, uh, it's about trying AI. Everybody, I'm not talking about this room, of course, but everybody out there is talking about AI. Very few of them are actually using AI, even fewer are understanding AI, and just a very small fraction of them uh is really doing AI for uh you know for real. So um the first step is trying. What we want is uh you know provide uh you know a way to you know present uh no code required experiments so that people can just uh you know consume, try in order to really appreciate what uh you know what AI is about. But then we want them to understand AI, right? And so we want to make available for them, for example, pre-trained models uh right in order to really appreciate the output of AI, what the AI is capable of. But that's just half of uh you know the journey. The next step uh is to take control, right? Uh we want them to be able to train AI, not necessarily starting from scratch, right? Uh starting from a pre-existing model, being able to either fine-tune it or improve the model, or uh provide them with all the tools that they need in order to feed their own data in order to train a brand new model that is unique to them from scratch. The final step, which is also the most important uh step, is actually make something with AI. Uh, Arduino was uh you know uh originally one of the companies that created the maker movements, right? And we think about makers as uh kind of you know uh geeks uh in a in a basement, right? Uh, but makers are people who make stuff, right? Uh everybody's a maker. If you make stuff, you are a maker, right? Whether you are a scientist uh you know at uh you know an ASA, uh you are an RD director at uh automotive manufacturer, or you're just someone at home that just want to automate uh your uh you know your shades. Um the thing about uh you know it's not just about the journey, it's about uh you know flattening the learning curve. Um it's about uh really scaffolding this journey in a way so that uh you know people can you know go along in the journey, keeping their confidence intact. Um there are a few pillars, right? And that's you know, now go into you know into the details of what we do in order to enable this vision, right? Uh uh it's all based on three areas. One is about hardware, right? Is really having uh a portfolio of uh you know AI and ML ready hardware that people can just uh get and buy without uh you know the need of building something from scratch. Um is about uh you know the openness of being able to support uh not just the models and the technology about AI that we provide, but an array and uh you know uh open uh ecosystem of uh you know models, AI frameworks, and so on. And then uh you know software and tools that are cloud native and can really support uh you know users and developers throughout the entire journey. So let's uh take a look at uh you know each one of them in detail. Um you know the funny thing is that every time uh that I talk to someone uh uh that knows the Arduino name, every single one of them uh thinks that Arduino makes only one product, the old 8-bit uh you know Arduino Uno microcontroller. This is actually just a part of the portfolio of products that uh at Arduino we actually make and sell every year. And if you see, we kind of spun across uh almost every uh MCU architecture that ARM has ever kind of published. And uh you know uh over the last two years we actually went over, right? Went beyond MCU and we actually started to ship uh you know MPU and even hybrid MCU and MPU devices. The thing is that uh the entire portfolio at Arduino is AI and ML capable. So it's not uh you know not anymore at the time of you know that tiny, you know, but great, uh, right, uh microchip uh microcontroller, right? Uh but it's really uh you know making available to everybody the hardware platform that is best suited to address the use case uh and to at the end of the day achieve the you know that vision uh that they have in mind. Um whether they're doing uh time and TinyML or they want something more powerful that is Linux-based, uh, we know we got it, uh, we got it covered. And uh this is not just uh part uh of uh you know the portfolio, but it's also you know what we have today. You will see in the next few months, in the next couple of years, uh, you know, the pace of innovation at Arduino is just accelerating, and you will see new hardware and new form factors that will be uh, you know, of course, more and more uh designed uh with uh you know AI, generative AI, computer vision, uh audio inference, speech to text uh you know in mind as uh you know core scenarios. Then uh software and you know, specifically framework. Um developers uh don't want and we don't want them to be locked uh in proprietary tools. They don't want and we don't want them to be locked into a single tool. And this is why this is just a very tiny example of some of the frameworks that we support that are you know out of the box. Some of them are kind of open source frameworks, some are you know commercial platforms, uh, some are you know kind of you know hybrid between the two. We create examples, we create native integrations, we create libraries so that those uh frameworks uh can be used and supported out of the box. And we foster a community of people that want to actually bring into the ecosystem new frameworks uh and new technologies. I was kind of looking at uh uh at the forum the other day, and there was someone that was kind of you know asking how they could uh you know add uh native support uh with a library to small language models uh in uh you know in the Arduino ecosystem. And uh a couple of our engineers are supporting them to be able to do so. And so the the key here is that uh uh you know if you if developers bring their own workflow, we amplify them. If they are looking for a framework and a workflow, we support them and we provide uh with examples uh in order for them to understand what's best, right? Uh whether it is uh you know an OpenMV you know framework to do computer vision of super tiny MCU-based camera, or they want something a little bit more refined on Gimpuls that they want to take a base model and train with their data set uh in order to do face recognition, right? And recognize one specific person and not just uh that that uh you know that that is a person. Then the third pillar, I would say literally last but not least, uh next generation cloud native developer tools. Uh let's face it, uh, we don't live uh off of native applications. I know about you guys, but you know, my life uh is running through Google Chrome or Microsoft Edge or Chromium, whatever is you know, the browser of your choice. Um and you know the the idea for us uh is really to you know provide uh uh and support and grow a vast array of tools that can be accessed. Uh no, it doesn't matter what type of computer, what type of machine, what type of power factor device you actually have. So from uh low-code to no-code tools uh to you know developer uh you know uh to you know the ability to uh deploy the code uh you know from a single device uh to you know deploying uh you know at scale with fleet management and over there updates if you have a fleet of devices that are um uh that are uh that are deployed in the in the field, but also having uh you know devices that once are connected can be connected to the internet, uh can send data to the cloud, not just for remote monitoring, predictive maintenance, and so on, but also to collect data and the same data that is then used to retrain the AI, to fine-tune AI models and so on. It's you know everything is uh you know brought together in not just in a single tool, right, but in an ecosystem of systems uh that are seamlessly working together and where developers can go pick and choose what they need and assemble them together. Uh part of these tools, uh, talking about AI, uh we are actually starting this year featuring uh you know something new, which is uh an AI assistant. And this is uh you know really AI doing AI, right? It gets a little bit recursive. Uh we introduced AI Assistant a couple of months ago, um, and is uh again embedded uh into the Arduino um uh toolchain on uh you know on the cloud. Uh Arduino AI Assistant is uh you know, I I like to call it more of a coach than a co-pilot, um, and it's something that developers can use uh not just to ask for you know code that they want to be generated, it's not just by coding, but is a fundamental tool for people to understand what they're doing. You can use the AI assistant uh, take a snippet of code off of uh Stack Overflow and have the i assistant explain what that code does. You can have your code that you wrote, it doesn't work, it doesn't compile. You can ask the AI assistant uh to analyze the code, understand what's wrong, and fix it and suggest fix, you know, fixes for you. You can have code, maybe you're not like me, a good developer, and uh you know you want it you know to be optimized, to consume less memory or to be you know faster. You can ask the assistant to rework your code in order to make it more efficient or to make it uh more suitable for a specific platform. Uh the idea is really, you know, think about think about it as a coach, uh uh you know, to level up uh your uh you know, embedded intelligence, uh, you know, if you if you if you will. Um last thing, uh, why should you believe what I'm saying? That's you know, if you look if you haven't asked this question, you know, you should, and I'm actually asking this question for you. Um well, you know, Arduino is an open community. You know, you can just take a look, go up there and uh, you know, just Google uh you know Arduino, Google Arduino Forum, look what people are asking and what they are saying about the result of their work. Uh go on Project Hub, see the projects that you know people are doing, just uh you know choose edge AI or AI, and you will find every project uh that you know people have published. But also take the word of top professional developers and companies that have standardized their innovation on Arduino. One example is BNW Group. We are quite big in automotive, every single automotive manufacturer in uh you know Western automotive manufacturer as you know is doing prototyping with Arduino, but BNW Group has standardized their innovation on uh on our technology. And uh they really rely on uh you know the framework and the uh you know the tools uh in order to standardize and uh you know really cut uh the you know innovation time from uh you know the drawing board uh to you know production. One example of the way that they actually implemented Arduino Net AI, you know, if you happen to buy our old source phantom, and I really hope that you do, um uh you will see that there are 1,000 LEDs uh in the ceiling, uh, uh which is something that they call the starlight headlining. And these lights are actually representing uh you know the stars in the sky of the location where you are. And the whole thing is actually controlled through an Arduino device, or actually to an Arduino uh design, uh, and is kind of you know using an Arduino framework with uh you know an ML framework that I'm not able to disclose. Is you know, they bought their own framework in order to being able to provide this type of experience uh you know to every single developer. So we want AI for everyone and we want it for real. This is not a slogan, this is a principle, this is a tenet uh that is what we operate with every day at Arduino. We mean it. We believe that AI must be inclusive, it must be educational, and most more important of all, it must be empowering. At the end of the day, what we want for you know the ecosystem up there, for every single one of you, is put you in the condition that uh if you can imagine, you should be able to build it. Thank you.