Simon Bae’s problem with the perfect plan

Simon Bae considered becoming an angel investor. He had founded two software businesses, both eventually acquired by 3D Systems. Writing checks for other people’s ideas was a plausible next chapter. He chose a third company instead. His explanation had a pleasing lack of ceremony: investing in himself had produced his best returns. Besides, he enjoyed solving problems.

That third company, Cupix, began in 2015. The problem it eventually settled on is familiar to anyone who has watched a building take shape: a plan describes an intention; the site keeps its own diary. Materials arrive. Installations change. Walls conceal yesterday’s work. By the time somebody needs to settle a question, the useful evidence may be somewhere behind the plaster.

Bae’s answer starts with a person walking around carrying a 360-degree camera. Cupix turns the footage into a navigable spatial record that teams can compare with designs and earlier captures. A complicated body of mathematics sits behind an ordinary activity. The person on site still has to walk. Bae has spent much of his career working on what the computer should do afterward.

A geometry enthusiast, with companies attached

Bae describes himself as a serial entrepreneur and a “3D geometry enthusiast.” The second description helps explain the first. His companies have changed customers and business models, but the recurring subject is physical reality: how to describe it, move it into software and let other people work with it.

He studied mechanical design at Seoul National University and earned graduate degrees in computational geometry there. His professional history includes postdoctoral research at Ricoh’s software research center in Japan and at MIT. Geometry, in this career, is less a school exercise involving obedient triangles than a practical question about how computers can make sense of objects that already exist.

His first venture, INUS Technology, developed Rapidform, software for working with 3D scans. Reverse engineering begins with the object and works back toward a usable digital model. Inspection brings another question: how closely does the object match its intended design? Bae was working on the relationship between an ideal description and a stubborn physical thing long before Cupix put buildings into the picture.

3D Systems acquired Rapidform in October 2012 for $35 million. More revealing for his later work is the technical lineage. Scan data became something engineers could use, and share with colleagues.

First the object. Then the people around it.

Bae’s second venture, VisPower Technology, formed in 2010, approached another part of the problem. Its product, TeamPlatform, let design, engineering and manufacturing teams manage projects and share data in the cloud. A digital model might be useful. Getting the right version to the right people was its own occupation.

TeamPlatform handled project information, tasks and reusable workspaces for companies dealing with many projects at once. The emphasis widened from describing an object to organizing the work around it. When 3D Systems acquired VisPower in August 2013, it said it would integrate TeamPlatform into its professional and consumer offerings.

Bae went on to serve as CTO of cloud strategy at 3D Systems. By the time he started Cupix, his career had covered two ingredients that would meet there: reconstructing physical things in three dimensions and helping dispersed people collaborate around them. Cupix’s combination makes more sense when viewed through both businesses. The camera supplies the evidence; shared software gives that evidence somewhere useful to go.

The house was the beginning

Cupix did not begin with its present construction brief fully drawn. Bae has described an early idea involving phone photographs, measurements and virtual interiors for homes. Real estate tours were part of the original opportunity. The company’s name joins “cubic” and “pictures,” a compact explanation of its starting material.

The decisive change was finding a setting where the record would be needed repeatedly. A property tour lets somebody inspect a space. A construction site changes between visits. Recording it again can answer a different question each time. Bae concluded that construction gave the technology a more useful job and redirected the business.

Cupix established its Silicon Valley subsidiary in 2017. Early demonstrations used inexpensive consumer 360-degree cameras and cloud processing to assemble photographs into a spatial scene. The ambition already included comparing captures across time. Construction offered a particularly literal reason to do that: a useful installation can become invisible as soon as the next trade finishes.

There was also a practical distinction in those early demonstrations. Quick measurements from camera-based reconstruction served documentation and communication; work demanding exact measurements could require laser scanning. Keeping the capture method suited to the task matters. A camera that people can carry easily has advantages, but a sensible workflow still needs to know what its measurements are for.

When the drawing loses the argument

Bae’s writing about construction returns to the gap between the designed building and the built one. A carefully prepared 3D model can lose relevance when work begins, changes accumulate and site documentation drifts into separate files. A drawing can remain perfectly legible while becoming an increasingly unreliable guide to the room in front of you.

CupixWorks connects captured conditions with building information modeling, or BIM. Teams can navigate the site remotely, compare installed work with the design and review earlier conditions. The extra dimension in a 4D digital twin is time. A room can be inspected as a place and revisited as a sequence.

The point becomes clearer with an ordinary construction question. Before closing a wall, a team records what is inside it. Later, someone needs to understand the installation. A dated, spatially located capture gives the discussion a place to begin. This is an illustration of the workflow, with none of the glamour ordinarily assigned to futuristic software. The wall has simply become less secretive.

Bae has argued for keeping that record useful after construction, through handover, operation and subsequent renovation. In that view, the finished building does not close the file. The next team inherits something it can inspect. A model that survives the handover can spare people the peculiar professional ritual of investigating work their own organization has already paid to do.

A smaller camera, a wider circle

The accessible camera is central to Bae’s approach. His earlier work involved software around specialized 3D scanning; at Cupix, the capture device could be a consumer product. Lowering the equipment burden makes frequent documentation easier to contemplate. Frequency matters when the subject can change before the next meeting.

Cupix has also worked with other software companies rather than asking a construction team to conduct every part of its day in one application. In April 2023, CupixWorks joined Bentley’s powered by iTwin program, connecting captured data and BIM for navigation and coordination. In March 2024, a partnership with NTT Communications opened another route into Japan.

In 2025, Cupix announced support for the Insta360 X5 camera. These connections describe an operating choice: improve the link between existing hardware, existing project information and the people trying to make a decision. The technology has to fit the working day. It is difficult to persuade a busy site team that the real purpose of its building is to demonstrate your software.

“Everyday, Everyone, Everywhere.”

Simon Bae’s ambition for Cupix

The move north of Austin

Bae moved to Williamson County, Texas, in 2024. Cupix’s American headquarters followed from Silicon Valley, with the move reported in April 2025. He pointed to business conditions and the available talent. The route there also involved Francis Pahng, Cupix’s AI executive, and connections through the Greater Austin Asian Chamber of Commerce.

It is a useful detail in a career easily summarized in acquisitions and algorithms. A multinational company still moves through people who know a place, can introduce a colleague and help a new arrival settle. Cupix’s current office list spans Round Rock, Texas, Seongnam in South Korea and Lavender Bay in Australia. Geography continues to matter to a business selling remote access to geography.

Bae has identified different business cultures and regulations as challenges of running across borders. His other leadership role gives that experience a wider audience. Elected chair of the Korea Proptech Forum in 2023, he was reappointed in February 2025 for a term extending to April 2027. His agenda includes international exchange and helping member companies grow. The founder trying to sell across borders also helps organize a forum for others doing it.

Teaching the record to answer back

Bae’s next ambition gives the accumulated site record a more active role. In a June 2025 interview, he described Compass, a conversational AI project being developed with AWS. A user could ask about progress or whether materials had been cleared. The system would need to connect the question with spatial records, drawings and project plans.

The sequence matters. Fluent language alone does not tell software which materials are in which room, or whether their presence is new. Bae’s argument is that AI needs spatial context before it can offer useful help with a physical site. The proposed assistant depends on the less theatrical work of capturing and organizing what happened.

By October 2026, he was presenting that direction as enterprise spatial intelligence. At a Seoul customer event reported on October 20, Bae discussed a camera-first approach, and Cupix showed CupixWorks 5.0 following its September unveiling at Autodesk University. Beta testing was scheduled to begin in November. That is a next step to be tested in customers’ working conditions.

There is a recognizable continuity here. Rapidform helped make scanned objects usable. TeamPlatform helped people collaborate around project data. Cupix brings a changing physical site into that conversation. The third company has a longer memory than the original property tour, and Bae wants that memory to support decisions.

His early Cupix ambition was expressed in three words: “Everyday, Everyone, Everywhere.” It is an exacting brief disguised as a friendly slogan. Everyday means the procedure cannot be a special production. Everyone means expertise cannot be the admission ticket. Everywhere means a business has to travel across borders as well as reconstruct rooms. Bae’s unfinished business is making all that complicated work feel ordinary enough that somebody will do it again tomorrow.