Interview with Elisenda Bou

Elisenda Bou-Balust holds a PhD in Engineering from the Universitat Politècnica de Catalunya in collaboration with MIT. She was co-founder and CEO of Vilynx, which was acquired by Apple in 2020. In 2022, Elisenda was awarded the Princess of Girona Award for her work in the field of Artificial Intelligence. She recently decided to co-found Cala, where she is also the CEO.

You began your academic journey in 2005, pursuing studies in Telecom Engineering at the Universitat Politècnica de Catalunya (UPC), where you graduated in 2010 with Highest Honors. During that same period, you also earned a degree in Electronic Engineering from the Universidad de Las Palmas de Gran Canaria, where you got an MSC Thesis. Moreover, you had time to get another MSC Thesis in a collaborative effort between MIT and the Technical University of Catalonia. I am deeply impressed by your accomplishments, and I am curious to learn more about your interests and reasons for such an exciting journey

Truth is, I wasn’t planning on studying Electronic Engineering at ULPGC. I originally went there for an exchange year, but I enjoyed it so much, and also landed what was, at the time, a dream job for me at Ideti, that I decided to stay and finish the degree as well.

During those years, I also had the chance to spend several summers (from June to October) at MIT, after being invited following my participation in a NASA Centennial Challenge. To be honest, the only thing I really knew about MIT back then was that Feynman had taught there, and as a huge fangirl, that was reason enough for me to want to go. That experience eventually led to my MSc thesis. It was a mix of curiosity, opportunity, and a deep excitement for working on things that felt a bit ahead of their time.

Later, between 2012 and 2016, you pursued and completed your Ph.D. on “system architecture for satellite constellations and on the exchange of resources among them” at the UPC. Could you tell me more about your doctoral research? 

That period coincided with the very early days of satellite constellations, when most of the focus was still on individual satellites rather than on networks of them. We realized there weren’t many studies exploring how to design and deploy constellations effectively, both in terms of their architecture and how they could share resources.

My PhD focused on how we could develop satellite architectures that enable efficient resource sharing, and the software systems needed to make that possible. I also worked on wireless energy transfer within those constellations, studying how power could be distributed from one node to others to keep the network operational.

I was doing most of that work at the UPC NanoSat Lab, which was an amazing place. It’s one of those rare environments where you go from theory to actual hardware: from designing and simulating satellites to seeing them built, tested, and launched. That hands-on, experimental approach really shaped how I think about research: it wasn’t just about models, it was about making things work in the real world.

Your first entrepreneurial experience was in 2011 as the founder and CTO of Zeptocode, a company that developed software for marketing campaigns. What were the main technical projects you managed, and what were the key achievements, challenges, and lessons learned from that time?

I have extremely fond memories of Zeptocode. It was something I built with my family, which made it really special. We worked together on different projects, and since marketing campaigns are naturally short-lived, it meant we could constantly start fresh and build new things from scratch.

I had been programming my whole life, but Zeptocode was the first time I was shipping things professionally: real campaigns for major brands, with huge traffic spikes and even hackers trying to take our systems down. It was both thrilling and humbling.

That rhythm of short, high-impact projects gave us the chance to experiment a lot, to try new technologies, languages, and frameworks just because we could. Zeptocode became a bit of a playground for us to explore the latest tools and models while delivering real work under pressure.

In 2012 you co-founded Vylnix (VILYNX). Out of curiosity, what’s the meaning of “Vilnyx”

Vilynx came from Vi (Video) and Lynx (from the Spanish sentence, tener vista de lince). Since we worked on multimodal content understanding (particularly around video, text, and audio).

Vilynx pioneered one of the earliest self-learning multimodal architectures for understanding and representing knowledge using Knowledge Graphs.  From your perspective, what was the primary technical breakthrough you brought to the market?

I think there were several breakthroughs. At that time, multimodal systems were still quite immature, so we had to build a lot of things ourselves: our own speech-to-text, computer vision, and NLP models. Knowledge Graphs were also quite unknown back then.

Our advances happened both at the model level and at the system level. On one hand, we pushed the boundaries of what these individual models could do. But more importantly, we figured out how to make them learn from each other, combining video, audio, and text in a way that enabled unsupervised learning. We were able to extract multimodal information from untagged data and use it to make each model better, which was quite new at the time.

None of that would have been possible without the team. It was a group of incredibly talented people, curious and bold enough to build things that didn’t yet exist.

Given that you managed the business from Barcelona while also working with teams in several U.S. cities, I imagine it required significant effort to coordinate and navigate cultural differences. What was your approach to getting these teams to work together as “one team”?

To be honest, I don’t think we ever felt part of different teams. We didn’t do anything specifically. We all cared so deeply about the same problems that it was enough to make it feel like one team.

And, what were the main use cases for Vilynx's clients?"

Video Summarization, Catalog Search, Search, and Personalization.. At the end, it was mostly information retrieval, what you would do today with RAG, but before LLMs existed.

In 2020, Apple acquired Vilnix. How did you feel about it?

I loved working for Apple. I wasn’t expecting to enjoy it that much.

A few months ago, you got back to the entrepreneur side and co-founded Cala, where you are also acting as CEO. What is Cala about?

Cala provides trustworthy information for agents that need external data. It’s an infrastructure platform where developers can connect their agents to get the reliable, contextual data they need to make decisions.

Our focus is on trust and explainability,  making sure agents not only have access to the right information, but also understand where it comes from and why it can be trusted.

Agents are definitely a hot topic, so I would like to better understand how Cala helps to optimize them

Our thesis is that trustworthy data leads to trustworthy AI. We ensure they have the right information so they can decide and automate reliably.

Who is the main target of Cala's services?

From startups to scaleups to Large Enterprises. Anyone building agentic workflows can leverage Cala to ensure they have the context they need.

A more philosophical question. Do you think LLMs are actually reasoning or, as some people claim, are just pattern recognition?

LLMs are, in the end, probabilistic models that learn patterns in sequences of tokens. I don’t think they really “reason”,  at least not in the way humans do. They’re incredibly good at recognizing patterns and reproducing structures that look like reasoning. The results can sometimes feel surprisingly close to reasoning, which is probably why this question keeps coming up. But I see it more as an emergent effect of scale and data rather than actual understanding.

Since March 2023, you have also been part of the leadership team at Friquifund, an association that helps vulnerable kids in STEM. What types of projects does Friquifund support, and which one are you most proud of?

Friquifund comes from a very simple idea: we’re a group of people who have been able to live well thanks to having studied STEM. And we believe STEM is still one of the few real social escalators left.

With that in mind, we try to help young, underprivileged kids who are passionate about science or technology to build a career in it. We support them through mentorship, access to resources, and funding for specific educational programs.

What makes me proud isn’t the projects themselves, but the youngsters who have gone through our programs and are now starting their careers in STEM. Seeing them take that step, often against a lot of odds,  is incredibly rewarding.

To finish the conversation, I would like to know your perspective on the Gender Gap in STEM. According to recent statistics, it remains significant, with women making up only 28% of the STEM workforce. Why do you think this is happening?

The real question is not just why this is happening, but why it keeps happening even though we believe we’re taking action to counter it. There are many factors that influence whether girls choose STEM careers. One of the main ones is the lack of references: they don’t see enough women in these roles to imagine themselves there. Another is how women in STEM are portrayed in media. When they do appear, they’re often shown as the “weird” or “asocial” girl whose entire identity revolves around being in STEM.

That’s problematic, because it doesn’t help young girls see themselves represented in a realistic or relatable way. We need more diverse, human portrayals of women in STEM: people who are brilliant, yes, but also multidimensional.

And what can we do about it? 

I think this is now about showing girls that STEM isn’t just about math or code: it’s a way to have an impact on the world, in whatever field they choose. Whether it’s health, climate, art, or social change, technology and science are the tools that let you shape things.

If we can help girls see STEM as a medium for creativity and purpose, not just a technical path, everything changes. It becomes about empowerment and expression, not just a career choice.