Paul Lwin and the Problem of a Thousand Boats

A golf swing is an unlikely place to begin a story about autonomous naval vessels. Yet there is Paul Lwin, several years before HavocAI, working on a way for a phone to watch a golfer and explain what went wrong. Golf already offers abundant opportunities for humiliation. Lwin wanted useful analysis to come without an expensive collection of sensors. The phone in your pocket would do the observing.

Today, the question occupying him has grown considerably: how does one person supervise a group of machines moving through the real world? His company, HavocAI, develops software for autonomous vehicles to operate together. Boats were the starting point. Air and ground systems have joined the picture. The ambition is measured in thousands of assets, although an ambition and a demonstrated fleet are different things.

The golfer and the military operator have an unexpected connection. Each needs something understandable from a technically complicated system. Following that connection through Lwin’s career reveals a founder interested in what happens at the point where engineering meets an actual person, with limited time and a task to complete.

The first uniform

Lwin came to the United States as a political refugee from Myanmar at ten. His family settled in California, and he became an American citizen in 2000. An embassy Marine was the first American he encountered. That meeting helped give military service a particular meaning for him: the uniform belonged to someone who had helped his family.

The entrepreneurial impulse arrived early, too. In high school, he and his brother started Freest, an advertising-supported dial-up internet service. They eventually approached a bank about funding. The discussion ran into a rather fundamental due-diligence discovery: one of the people pitching the business was still in high school. Even a promising business model occasionally has to wait for graduation.

Lwin later reflected on meeting other Burmese refugees during his travels. Those encounters sharpened his appreciation of the opportunities he had received. His desire to keep serving, he explained while entering business school, grew from that comparison. This was a personal account of obligation, expressed before HavocAI existed and before autonomous boats became his public calling card.

It gives the later career choices a useful context. He would move between military institutions, universities, and businesses. The recurring purpose, in his own telling, was to use what he had learned to make a contribution. Entrepreneurship offered another route for doing that.

A cockpit is an unforgiving classroom

The Naval Academy gave Lwin his undergraduate education in aerospace engineering. He went on to serve as a Naval Flight Officer, including combat missions over Iraq and Afghanistan as an EA-6B Prowler weapons systems officer. His work concerned electronic warfare: a technologically demanding mission conducted aboard an aircraft with a long service history.

He has described that experience as an early reason for his interest in defense innovation. Modern digital tasks and older equipment had to coexist. For the person using the equipment, the distance between what technology could do and what an institution could deliver was more than an abstract management problem.

Flight testing brought him further into development. A graduate of the United States Naval Test Pilot School, he flew in more than 25 aircraft types. His education also included master’s degrees in aerospace engineering at Maryland and computer science at Johns Hopkins. The combination put him on both sides of a familiar exchange: the operator reporting a problem and the engineer trying to understand it.

His early company, Curiosity Innovations, carried that interest into software. Its stated focus was enterprise applications, efficiency, and better user experiences. One project, TripNos, brought invitations, schedules, polls, and shared expenses together for group travel. Anyone who has tried arranging a trip by an expanding email chain can appreciate the scale of that particular enemy.

There is an instructive modesty to the problem. The users already know where they want to go. The software has to help them coordinate. Lwin was working on tools intended to make complicated activity easier to organize, well before the activity involved fleets.

The golf swing in the middle of the story

At Yale School of Management, Lwin added business training to his engineering background. He had entered the MBA program in 2018 while developing Curiosity Innovations. Finance and the civilian business world presented things he still needed to learn. Going back to school was a practical response to that gap.

By 2020, Yale had named him an Entrepreneurial Fellow for Steady Heady. The venture used machine learning to analyze golf swings recorded on mobile devices. Instead of asking players to interpret a pile of measurements, it aimed to show the swing with a mistake highlighted and provide immediate feedback.

“Our main goal with Steady Heady is to make sports technology more affordable and accessible.”

Paul Lwin, on his golf technology venture

The motivation was specific. Growing up, he said, his family could not afford sports camps or expensive equipment. He wanted aspiring athletes to have access to tools usually associated with private instruction. The product’s promise concerned access as much as analysis: useful information through a device people already owned.

At JoyMotion Golf, where his team biography called him “Coder-in-Chief,” Lwin worked alongside golf instructor Joy Bonhurst and fellow Yale MBA Frank Shaw. Bonhurst brought coaching expertise; Shaw brought finance experience. Lwin’s biography also connected the work to data-based learning platforms he had helped develop for aviators.

The pattern is visible without pretending that a golf lesson prepares someone for every problem in robotics. Domain expertise matters. So does the design of the feedback. A system that produces information still has another job to finish: helping its user understand what to do next.

Twelve boats make a different argument

In 2023, Lwin and his friend Joe Turner spent weekends experimenting with autonomous boats. Their materials included canoes, dinghies, and commercial components. It was the sort of beginning that looks agreeably improvised when viewed from the far side of a funding announcement. At the time, it was practical work on things that needed to float, move, and respond to code.

Turner brought his own naval and manufacturing experience, including work on autonomous underwater vehicles. Their experiments coincided with a growing Pentagon interest in large numbers of autonomous systems. Lwin heard customers asking whether companies could build thousands. A prototype answered only the first part of the request.

HavocAI began in January 2024. In an early public account, Lwin said the team built twelve vessels and took them to a defense exercise for users to try. Starting with a dozen made the argument tangible. Repeatability and coordinated operation were part of the proposition from the beginning.

The manufacturing choice was equally consequential. Havoc could work with people who already built boats and devote its effort to the software that made vehicles operate together. Rhode Island supplied an established maritime ecosystem. Hull construction, communications, autonomy, and the operator interface could each receive attention from people equipped to work on them.

Coordination is where the arithmetic becomes demanding. One operator cannot give a thousand vehicles a thousand streams of individual instructions. Havoc’s approach lets the person assign objectives and supervise execution while software handles more of the movement and collaboration. Increasing the fleet makes the interface a central engineering concern.

“If the software isn't this simple, it’s a science experiment,” Lwin says. It is a pointed test for a product: the demonstration must make sense to the intended user. Complexity has a way of seeking refuge in a training manual. Lwin’s stated preference is to remove some of it before the manual becomes necessary.

A handshake, a shipyard, a wider map

In September 2024, Havoc announced an $11 million seed round led by Scout Ventures and co-led by Trousdale Ventures. The money supported production and hiring. It also moved Lwin into the familiar founder’s task of turning a technical approach into an organization capable of delivering it repeatedly.

By May 2026, the manufacturing model had a public, local expression. Lwin and Senesco Marine president Ted Williams announced a partnership at Senesco’s Rhode Island shipyard. Both executives were military veterans and Naval Academy graduates. The plan paired Senesco-built vessels with Havoc’s collaborative autonomy software for defense and commercial applications.

That month, Havoc also announced a $100 million Series A, bringing its reported total funding to approximately $200 million since 2024. Acquisitions of Mavrik and Teleo extended its scope into aerial and ground systems. The financial milestone enlarged the assignment: more kinds of vehicles, more integration work, and more people depending on the same basic idea of coordination.

September brought two further examples. An integration with Rune Technologies connected logistics planning to Havoc’s autonomous aircraft. The companies said they had completed an end-to-end resupply mission during an August field exercise at Fort Hood. The work put movement in service of a concrete need: getting supplies to a unit.

A separate partnership with Taiwan’s CSBC proposed licensed co-production of autonomous surface vessels. The intended division of labor was clear: Havoc’s autonomy software, with platforms produced and sustained in Taiwan. It extended the manufacturing partnership model beyond Rhode Island while leaving delivery and long-term performance as work still to be done.

Lwin has described his larger purpose in terms of deterrence and protecting military personnel. Those goals carry obligations that a funding round cannot settle. Systems have to work, users have to understand them, and useful performance has to survive outside a demonstration. His career has repeatedly brought him back to that encounter between a person and a machine.

The phone watching a golf swing now seems a little less incidental. There, too, the task was to turn technical capability into something someone could actually use. Lwin has changed the machinery, the setting, and the stakes. The person holding the device still has a job to do.