You may not know the Aechelon name, but if you’ve flown for the U.S. Air Force, Navy, Marine Corps, Coast Guard, or Special Operations community, there’s a good chance you’ve trained in one of our simulators.

That’s probably not where you’d expect a company that started making visual effects for movies to end up.

About 27 years ago, I ran a company focused on special effects, and one day somebody came to us and asked, “Can you guys build us a flight simulator?” We looked at each other and thought, what the heck is a flight simulator? But we built one anyway, then kept building them better and better until eventually we met customers who needed high-quality simulators, and they told us, “We love this.”

We transitioned away from special effects, focused entirely on simulation, and never looked back.

Today, we generate what pilots see out the cockpit window, or what their sensors see when they’re looking through radar, infrared, electro-optical cameras and other onboard systems. Unlike companies focused on one prime contractor or a single aircraft platform, we’ve always built products for the broader market and ecosystem. There’s probably not a U.S. military pilot flying today who hasn’t touched one of our systems at some point. That reach matters now for a new reason.

For most of those 27 years, we thought we were in the business of training pilots. Then we realized the infrastructure we’d built to train pilots could also train machine learning systems. The environments, sensor models, and data we’d spent decades building for pilots turned out to be exactly what autonomy needed too.

That insight is the heart of why Shield AI and Aechelon fit together. The future will not be a world of only human pilots or only autonomous systems. It will be a hybrid environment, and both sides need to train in the same world environment.

Humans and Autonomous Systems Must Train and Fight Together

Long before there was ever a conversation about an acquisition, our teams were already working together. At Aechelon, we’ve spent our existence trying to help keep pilots safe. Shield AI exists to protect service members and civilians with intelligent systems. We realized we were solving the same problem from different directions: What resonated with us most was that we weren’t changing our mission by joining Shield AI so much as extending it.

The best way to keep a pilot safe isn’t just giving them better training. In many cases, it means not sending them into the highest-risk part of the mission first. Let autonomous systems take on that risk while humans supervise and make key decisions. The more we can keep people out of harm’s way, the better it’s going to be for everybody.

That naturally leads to the next challenge, which is that I don’t think we’re going to live in a world that’s only human or only autonomous. You can already see that with Collaborative Combat Aircraft, and it’s only the beginning. That means we need to train humans and autonomy to operate together. Simulation is much more than graphics; we’re building a representation of the world. Human pilots train in that world today, and autonomous systems are going to have to operate in that same world tomorrow. They need a common representation of the world and a place to learn together.

Autonomous systems need to learn how human pilots fly, while human pilots must learn how to work with autonomous teammates and how to defend themselves against autonomous threats. These are skills that don’t really exist today because this space is still new, but they’re becoming essential.

Closing the Data Loop Will Accelerate Autonomy Development

Our Shield AI Co-Founder, Ryan Tseng, talks a lot about “closing the loop,” and I think that’s exactly the right way to think about why bringing together Shield AI and Aechelon makes so much sense.

Like pilot training, there’s a cycle for training machine learning systems. You simulate and gather data. You train. You test. You deploy. Then you gather even more data, fine-tune the models, and close the loop. Every time you go around that loop, the system gets better.

What excites me is that we can now connect high-fidelity simulation with defense-grade autonomous systems in that loop. Everybody has heard that autonomy needs three things: compute, data, and a platform where the model can actually operate. Hivemind is the brain. Aechelon brings the data. Shield AI’s aircraft and partner platforms provide the operational environment. Having those three things together in one closed loop that is continuously improving is what makes this combination so powerful.

Better data produces better autonomy, and better autonomy produces better data.  Information from Hivemind and operational platforms feeds into the simulation, which generates even better training data. Everything improves everything else. That’s the flywheel.

Speed and Scale Will Drive Exponential Training Outcomes

When simulation and autonomy are fully integrated, speed becomes possible. Simulation lets us scale learning events in a way that’s impossible in the physical world.

Autonomy is improving at an exponential rate, and humans are really bad at understanding exponential change. At first, change feels incremental. Then you close your eyes for a second, and suddenly everything has changed. We’ve already watched that happen with large language models. I think autonomy is going to follow exactly the same path.

In a fully synthetic environment, every input the autonomous system receives is simulated. We generate radar returns, infrared imagery, and electro-optical feeds that the system would see in flight. From the system’s perspective it’s operating in a realistic mission environment, even though it’s entirely virtual.

We’ve actually done this before. We took an operational Navy system, put it into a simulated environment, and convinced it that it was flying because every sensor input matched what it would see in the real world.

That changes the pace of learning. If you’re training with a human instructor or a real aircraft, you can only go so fast. You’re limited by aircraft availability, weather, maintenance, schedules, and cost. But because the environment is synthetic, you can now scale it. Instead of training Hivemind with one hundred missions, we can train it with one hundred million missions. That’s how you make autonomous systems improve very quickly. That’s how you expose it to edge cases that might take years to encounter in the real world. That’s also how you build systems pilots and operators can actually trust.

Strengthening Deterrence Through Autonomy

As we begin this new chapter together, my biggest measure of success isn’t simply seeing more autonomous systems in the field. It’s seeing people’s trust in autonomy continue to increase. The more data we generate, the more training we do, and the more we fine-tune these systems, the safer and better they’re going to become. Seeing that trust grow is how I’ll know we’re fulfilling the potential of what we’re building together.

I’ve always believed that the best war, by far, is the war that’s never fought. Nobody likes war less than the people who actually have to go to war. That’s why I think deterrence is such an important part of this conversation. Strong, trustworthy autonomy doesn’t just help if conflict happens. It can help prevent conflict in the first place because credible capability changes the decisions other people make. Our responsibility is to make sure it’s developed by the people who are trying to do the right thing—that it reflects the values we’re trying to defend, that it’s safe, and that it helps reduce unnecessary loss of life.

For decades, Aechelon built the worlds where pilots learned to fly and fight. With Shield AI, we’re bringing those worlds together so that humans and autonomous systems can learn together, improve together, and operate together. If every improvement we make—from better simulation to better data, to better autonomy—helps reduce collateral damage or helps even one more service member come home safely, then I think we’re accomplishing exactly what our teams set out to do.

About the Author:  

Ignacio (Nacho) Sanz-Pastor is the General Manager of Aechelon at Shield AI, where he oversees Aechelon’s product and customer roadmap. Nacho co-founded Aechelon Technology, Inc., in 1998 and served as its CEO until 2026, when the company was acquired by Shield AI. For more than two decades, he led the company’s growth and innovation, guiding Aechelon as it reshaped the visual and sensor simulation, ISR, and autonomous AI data generation industries.