back to MWC after 5 years: AI, aerospace, and new beginnings

¡Hola! After 5 years, I’m back at MWC26/4YFN and Talent Arena. I’m genuinely excited to be there to see what companies are building and, especially, to gauge the massive impact AI and robotics are having right now. I’m hoping to get inspired and apply these state-of-the-art (SOTA) breakthroughs to the aerospace industry. Honestly, I think there will be a ton of next-gen news and surprises.

Taking advantage of this context and the credits Microsoft is providing, I’ll be putting together a project-focused course called “AI for Aerospace”. In it, I’ll dive into ML/AI algorithms applied to the aerospace industry. It’s exactly the kind of course I wish I had when I first got into ML/AI algorithms during the COVID pandemic. How time flies!

From MNIST to Multimodality

Speaking of machine learning algorithms, it’s wild how much the industry has shifted since 2020. I never would have guessed that using the MNIST dataset to train a basic model for recognizing handwritten digits would evolve into something as massive as large language models (LLMs). In simple terms, they’re just predicting the next token from a prompt. Similar techniques also gave us multimodal models, which felt unthinkable back then.

In just 6 years, things have changed so fast that it’s almost impossible to keep up. It motivates me, but it also scares me a bit. I get the exact same feeling I had when I was designing an eVTOL MVP with my colleague Ferran Morales. We used to watch announcements from companies like eHang doing tests, and then boom, practically overnight they were starting operations for airworthiness certification. Then there was Lilium, suddenly announcing collaborations with airlines like Lufthansa, or planning to build vertiports in Florida.

The hype was thrilling, though unfortunately, it deflated as many eVTOL companies went bankrupt, including the German Lilium, which I really loved. It was an elegant piece of engineering. I genuinely hope the AI industry doesn’t suffer the same fate. I have high hopes that AI will continue to transform the aerospace sector.

What’s Next?

While I hope the AI industry avoids the burst bubble of the early eVTOL space, the best way to ensure real value is through hands-on education. That’s why, down the road, when I map out the roadmap for my “AI for Aerospace” course, I’ll share it with you all. I’ll explain my own journey going from MNIST datasets to Copilot, MCPs, agents, and skills. They have seriously become fundamental tools for my daily automations.

I’ll also take the opportunity to share my perspective, including what I actually do like about AI tools like Cursor, Windsurf, and Claude Code. I want to emphasize what I consider truly important for software development: good practices, solid design, and most importantly, how we can integrate AI into workflows while aligning with the aerospace industry’s strict regulations.

In my specific case, working with Aviation Ground Systems CNS/ATM (DO-278A/ED-109A), aviation is a very slow-moving sector. We still rely on things written before I was even born…

Making the entire aviation system safe requires tons of paperwork, standards, and strict regulations.

My hope is that in the future, with the help of AI, we can automate certain tedious tasks, or at least have tools that give engineers superpowers to boost their creativity. And that’s the real key. I don’t believe AI will replace the engineer’s work. I really love the concept of the “Augmented-Engineer”.

Anyway, leaving you with that thought. I’m hoping to find some truly mind-blowing stuff at MWC26/4YFN. Catch you there!