For a while, Mikkel Wad Thorsen and the early Teton team were practically living at Nykøbing Falster Hospital. It was not the glamorous version of startup life, with whiteboards and cold brew under flattering pendant lamps. It was the useful version. They gathered data, watched people work, built product, discovered what they had misunderstood, and built again. Thorsen later wrote that they got so close to the staff they were almost becoming care workers themselves. The point was not cosplay. It was comprehension.
That period supplies the most revealing line in Thorsen’s founder biography. Before Teton had a sizeable team, international plans, or a Danish supercomputer helping train its models, it had proximity. The company’s founders wanted to apply deep learning and computer vision inside care environments. Yet their first job was decidedly analogue: learn how a shift actually moves.
Care work is made of overlapping clocks. A staff member may be helping one resident, listening for another, remembering a third, and documenting something that happened fifteen minutes ago. Most software adds a fresh demand to this choreography. Open the screen. Enter the event. Confirm the box. Teton’s wager was that a system in the room could observe activity, translate it into anonymized information, and offer an alert or record when it mattered. The interface would begin with attention rather than input.
The education of an attentive founder
Thorsen’s training makes the shape of the company easier to understand. At the Technical University of Denmark, he completed a master’s degree in Design and Innovation in 2019, focusing on computer science, digital product design, and product management. His studies also took him to South Korea for robotics. He later worked as a teaching assistant at DTU and as a visiting researcher at the University of California, Berkeley.
This is not the straight-line biography of a model researcher who happened upon a market. It is the education of someone interested in the seam between technical capability and human use. Design asks who must live with a system. Product management asks what makes it survive contact with budgets and habits. Computer science asks whether it can work at all. Teton sits where those questions quarrel.
In 2020, Thorsen co-founded the company with Esben Klint Thorius, who became chief technology officer. They started with a broad mission: use emerging technology to improve daily work in hospitals and care homes. The first product, called Nightingale, took roughly two and a half years to build. A name borrowed from nursing history carried a very modern bundle of hardware, computer vision, software, alerts, and implementation work.
A camera with manners
The visible object in Teton’s system is a sensor mounted in the room. The more important object is an argument about privacy. A conventional camera records an image for a person to review. Teton’s system is designed to process activity into an anonymized representation rather than operate as a window for casual viewing. The company describes its wider ambition as a foundational data layer for the point of care: a persistent understanding of events, trends, and workflow that staff can use without replaying a person’s private life.
That distinction is commercially practical as well as philosophical. A technology that enters bedrooms, hospital rooms, and care facilities cannot treat trust as a paragraph in the terms of service. It must make restraint part of the architecture. Thorsen’s product language repeatedly returns to making the system easy to implement and useful to staff. The cleverness has to arrive dressed as a colleague, not an inspection.
In a difficult workflow, adoption is part of the product. Installation, integration, onboarding, and support are not chores left behind after the demo. They are the route by which the demo becomes ordinary work.
Thorsen was explicit about this as early as 2022. However smart a system might be, he argued, it creates value only after successful implementation. Teton committed to assisting with integration, setup, staff onboarding, and support. This is founder language with the showroom polish removed. High-stakes software rarely fails because its launch video lacked kinetic typography. It fails in the handoff between the promise and Tuesday afternoon.
From snapshots to sequences
By late 2024, Teton had moved beyond its first deployments. The company employed about 40 people and was working across four Danish regions and twelve municipalities, with collaborations in England, Switzerland, and the United States. Then came Gefion, Denmark’s AI supercomputer, backed by the Novo Nordisk Foundation and Denmark’s Export and Investment Fund and built with NVIDIA technology.
Teton was selected as an early startup user. The compute opened two lines of work. One was to help models understand activities unfolding across longer spans of time. The other was to explore digital twins of hospitals and care homes, allowing teams to simulate operational changes. Thorsen explained the first with a modest human comparison: people can often look a little ahead and sense what may happen in the next thirty seconds from someone’s behavior. Better models might bring software closer to that kind of contextual anticipation.
The shift sounds small, but it changes the category of the product. Detecting an event is a snapshot. Understanding a sequence is a story. The latter can reveal patterns, support planning, and suggest what deserves attention before a routine becomes an incident. Teton began talking less like an alert company and more like a predictive intelligence company.
Capital for the crossing
In September 2025, Teton announced a $20 million Series A led by Plural, which had also led its earlier seed round. Bertelsmann Investments, Antler Elevate, Nebular, and PSV Tech joined the financing. The company put the money behind a clear set of tasks: deepen its European presence, enter the United States at scale, and grow the team.
A large round gives a founder permission to accelerate. It does not decide what deserves speed. Thorsen’s public framing remained consistent: move care from reactive response toward predictive support. In his telling, the economic value follows from the operational one. Earlier information can help teams plan staffing, reduce emergency responses, and make each intervention more deliberate.
The American expansion makes Teton’s implementation philosophy newly important. A product born beside Danish workflows now has to accommodate different operators, buildings, records, expectations, and procurement habits. Software may cross the Atlantic in a deployment pipeline. Trust takes the slower route. Partnerships and integrations will determine whether Teton feels like one more dashboard or becomes part of the building’s working memory.
The work behind the work
Thorsen’s public personality is most visible when he talks about teams. In an early hiring post, he promised a strong sense of ownership and actual ownership. More recently, a collaborator recalled Thorsen describing his job as crafting an engine of work where people thrive because they are succeeding in their dream role. The language is earnest in its faith that an organization can be engineered for both output and decency.
He also seems delighted by signs that the product has escaped the sales deck. In a recent post, he highlighted a care job advertisement that described a modern workplace with Teton installed in every room. To him, the appearance of the technology as a recruiting attraction was worth celebrating. A staff member choosing an employer because the tools might make the day better is a different kind of product review, and possibly the more honest one.
The arc from hospital fieldwork to international expansion can make Teton look inevitable in retrospect. It was not. The first product took years. Deployments required patience. The technology had to watch without making people feel watched, alert without becoming noise, and document without turning context into bureaucratic confetti. Those tensions still exist. Scale will sharpen them.
Thorsen’s useful idea is that the answer begins near the work. Stay long enough to notice the small frictions. Build the system around them. Treat implementation as design. Use bigger models when the question truly needs them. Then return to the room and see whether Tuesday afternoon has improved.
There is a pleasant irony in a computer-vision founder whose distinguishing habit is listening. Teton’s machines are being trained to notice patterns across time. Its CEO learned the same lesson first: attention compounds. A shift becomes a dataset. A repeated nuisance becomes a product requirement. An ordinary day, observed carefully enough, becomes a company.