When I connected for this interview, the team at Unbound Autonomy was in the middle of yet another hardware iteration. Their wearables sit at the intersection of two worlds that rarely move at the same speed: frontline requirements that change week by week, and manufacturing cycles that still run on lead times, component constraints, and slow hardware iterations.

Unbound Autonomy is building a compact battlefield wearable that passively captures operationally relevant speech and sound events, structures that information, and relays it to commanders.

The aim is not to add another screen, but to reduce the manual friction that still defines much of battlefield reporting, where updates are passed verbally or written down by hand.

In this conversation, the company’s co-founder, Vytautas Mikalainis, explains how early conversations with Ukrainian operators led to a pivot away from an AI “assistant” concept toward a tool that automates the unglamorous but critical work of battlefield coordination.

He also shares why the team ultimately had to build custom hardware, where they are seeing early traction in Lithuania and Ukraine, and what it takes to make on-device AI work under strict power and compute constraints.

Who Are You and What Are You Working On?

I was born and raised in Vilnius, Lithuania. I studied software engineering and spent more than seven years working as a software engineer before gradually transitioning into more management-oriented roles.

Alongside that, I have always been very interested in geopolitics. Since 2014, and even earlier events like Russia’s war in Georgia in 2008, it was hard not to see patterns repeating. Over time, I felt a growing need to contribute in a tangible way, not only by discussing problems but by helping solve them.

I joined the Global Shapers community, which is affiliated with the World Economic Forum, and worked on social initiatives in Lithuania. When Russia launched the full-scale invasion of Ukraine, we also supported refugee-related efforts locally, helping people find work, housing, and childcare.

Eventually, I left my job and joined an entrepreneurial program designed to help people find co-founders and build new ventures. That is where the path toward defense began.

Where Did the Idea Come From?

I met my co-founder, Nojus Kybartas, through that program. Initially, we worked on something unrelated to defense, but we lost momentum and began to stagnate. At the same time, we kept having long conversations about what was happening in Ukraine and how technology was changing the battlefield, especially drones and electronic warfare.

We were both very sensitive to what was happening there. Nojus had studied autonomous systems and had also personally supported Ukraine through donations.

When we decided to start something together, we first explored an idea that sounded strong on paper: an AI assistant for soldiers, something that would listen, analyze, and give advice in real time through a helmet.

Then we went to the first European Defense Tech Hackathon in Munich and pitched the idea to a Ukrainian fighter. His reaction was immediate: not only was the idea imperfect, but it was also fundamentally unrealistic.

He told us the frontline is too loud, too chaotic, and too human for that type of “AI voice in your ear” concept. Trust, communication, and human judgment matter more than an automated advisor telling you what to do.

That feedback forced us to reset.

How Did You Continue From There?

The turning point was a contact Benjamin Wolba shared with us, a Ukrainian commander near the frontline. We called him and had a blunt conversation about how things actually work.

He described a shockingly manual system. Communication happens over handheld radios, updates are passed person to person, and someone often writes key information down by hand in a journal.

Positions, health status, resources, changes in the situation, a lot of it is handled in ways that do not scale under pressure.

We asked whether it would help if this could be automated. He said yes.

During our next European Defense Tech Hackathon in Paris, we built a prototype that transcribed radio communication and structured it into a system, eliminating the need for manual note-taking. That became the foundation.

Later, we joined Scalewolf, which invested in us early and helped connect us to the right people.

We tested with a Lithuanian reconnaissance unit and kept iterating based on what units asked for, not what sounded clever in a pitch.

What Are You Building Today?

We are building a wearable device for soldiers. The goal is to passively capture military-relevant conversation and specific sound events, then structure that information so commanders can understand what is happening without relying on handwritten notes or fragmented radio updates.

It is not about listening to everything. It is about capturing what matters. We filter out irrelevant chatter and focus on operational content and sound events. We also work on the “cocktail party problem,” aiming to pick out relevant signals even in noisy environments.

The system includes wearables and a router unit, basically a rugged case that collects data from the wearables and displays soldier status updates.

Why Did You Decide to Build the Hardware From Scratch?

We tried not to.

We looked at existing boards like Arduino and tried using larger, off-the-shelf options like the Raspberry Pi. Each path had a blocker. Raspberry Pi requires too much power. Some Arduino options were not flexible enough for our needs.

Our requirements were not just audio. We also needed a GPS for training use cases and a way to transmit data, which meant a radio module. Off-the-shelf solutions did not cover the full set.

So we hired a PCB designer and went through multiple iterations. Eventually, we built a custom board.

The hardest part is not just building the hardware. It is ensuring on-device AI runs on a small processor while keeping battery consumption low. We are working with constrained compute and strict power limits, which force careful optimization.

Where Are You Seeing Traction?

We are still at the stage where validation and testing matter more than closing revenue.

We have strong interest from Ukrainian units and battalions who want to test the system in the coming months. In Lithuania, we have interest from the Lithuanian Riflemen’s Union and from units connected to prior testing, including the reconnaissance unit we worked with after Scalewolf.

We also receive feedback from people with experience in Israel and the United States. One of our advisors served in the US Navy SF, and his feedback has been valuable for shaping the product and its long-term vision.

Our near-term focus is to finish our “homework” in Lithuania and Ukraine, prove the product in real environments, and then expand testing into other countries, such as Poland, and potentially beyond.

If You Launched Another Battlefield Electronics Project, What Would It Be?

I would focus on something deployable, devices you can easily place near the frontline to collect useful metadata.

That could include radio data, acoustic data beyond what a human ear can reliably catch, vibrations, and patterns that help you understand what is happening. Not necessarily tracking troop movements directly, but collecting signals that, when analyzed together, can generate valuable insights.

I think low-cost deployable sensing can be very powerful if it is designed for real operational needs.