The office had a wood-burning stove, a stack of firewood and windows that did a rather casual job of keeping the warmth inside. On the floor sat a 3D printer. Mikkel Wad Thorsen and Esben Klint Thorius were building their startup above a former Copenhagen shipyard, in premises shared with artists and other creative tenants. Artificial intelligence had acquired a decidedly practical landlord.
That was the setting in 2020. Thorsen, now Teton’s CEO and co-founder, was already wary of a familiar engineering temptation: becoming so attached to a technology that finding a problem for it becomes an afterthought. His warning was directed at people like himself. Technical enthusiasm was useful. It could also give a bad idea a surprisingly long lease.
He had a comic image for the danger of hanging on indefinitely: failed rock stars, aged 45, still trying to invent the next AC/DC. The founders would give their venture a serious attempt, he said, and do something else if it failed. There was ambition here, certainly. There was also an exit from the mythology of the founder who must believe forever.
A product education, in three parts
Thorsen’s public education record at the Technical University of Denmark covers 2017 to 2019. He describes his focus there as computer science, digital product design and product management. The combination is revealing: how a system works, how someone encounters it, and how it becomes something an organization can actually deliver.
Those subjects have different habits of attention. A computer scientist can look for a better method. A designer has to consider the person using it. A product manager must decide which improvements deserve the next week of a team’s time. Put them together and the interesting question becomes where the technical possibility meets an everyday obstruction.
That is a useful way to read his career, rather than treating the letters AI as a complete explanation. His training placed the product alongside the computation. A model’s capability can be impressive in isolation; usefulness requires the surrounding decisions to make sense as well. The customer experiences the whole thing, including the awkward parts that never appear in a demonstration.
The money arrives after the relationship
In April 2023, Thorsen announced a $5.3 million funding round led by Taavet Hinrikus and Plural, with Finn Murphy participating. His public thank-you was expansive. It included the team, early collaborators and the people who had backed the company before its proposition was easy to explain. The announcement sounded like someone finally able to invite more people into a workshop.
Funding changed what the business could attempt. It also introduced a new test of the founder’s earlier discipline. A small team has obvious limits; fresh capital creates more possible projects, hires and markets than anyone can pursue at once. Choosing becomes part of the work.
Thorsen’s connections to Hinrikus and Murphy would continue beyond that announcement. By September 2025, Plural was leading a $20 million Series A, joined by Bertelsmann Investments, Antler Elevate and Nebular, with follow-on investment from PSV Tech. The company said it would expand across the United States, deepen its European operations and grow its engineering team.
Separate announced rounds. Funding amounts are not valuations.
At that second round, Teton reported that annual recurring revenue had grown thirteenfold since seed. It is a company-reported figure, with a specific starting point, rather than a measure of Thorsen’s personal wealth. Its relevance to his story is more straightforward: the venture had moved from persuading people to try something to managing the consequences of people buying it.
What a larger company owes its customers
Growth has a less photogenic side. Devices must remain connected. Updates must arrive intact. A release needs a way back if something breaks. These are excellent subjects for an engineering meeting and rather difficult subjects for a celebratory founder photograph.
In June 2024, Teton software engineer Luís Silva described the team’s work on Agent Smith, a tool for managing its distributed devices. Existing products had presented obstacles, including large software payloads and difficulties integrating NVIDIA-specific code. The team built its own system in Rust, then revised the design as it learned what remote devices demanded.
One of its decisions was to use established Debian packages and systemd, the service manager, rather than maintain every mechanism itself. The detail belongs in Thorsen’s company story because it shows the balance a founder has to sustain: giving engineers room to build while accepting that dependable, existing tools can be the sensible answer.
Silva wrote the engineering account. The work should stay attached to the people who did it. Thorsen’s role as CEO sits around that work: leading a business whose promises depend on a team making thousands of small technical decisions hold together.
“We solve them by shipping products that work, today.”
Teton’s published approach to problems
An unusually useful thing to give away
By October 2025, another Teton engineer, Alvaro Molina, described a fleet that had grown from hundreds of devices to thousands. The company was publishing Smith as open-source software. The tool handled deployment, monitoring and updates; other engineers could inspect it and contribute.
The rationale was commercial as well as collaborative. Fleet management was necessary to the business, but it was not the product customers were buying. Letting customer IT teams examine how it worked could build trust. Keeping every piece of infrastructure secret would offer a much less certain advantage.
For a company selling AI, this is an instructive allocation of attention. Some of the most consequential work happens beneath the feature that gets the meeting booked. A customer may never ask for a fleet manager by name. They will certainly notice if the equipment stops working.
It also complicates the familiar picture of a founder accumulating proprietary technology. Under Thorsen’s leadership, Teton could build something important to its own operations and still make it available outside the company. The practical question was whether openness helped the business and its users. Ownership alone was an incomplete answer.
Even the brain gets rewritten
A separate account of Teton’s software gives this practical streak a rather splendid name: Deep Snake Brain. The team replaced a complicated inference pipeline built from several C++ components with a simpler Python application using NVIDIA DeepStream. It reduced the number of repositories from three to one and the codebase by more than 40 percent.
The team reported that Python performance, despite its initial concerns, was not the bottleneck. Most of the time was spent running the model. Simplifying the surrounding software made development and testing easier. A choice that might sound less impressive in a programming-language argument made better sense in the actual system.
There is a management lesson in that episode, although it belongs first to the engineers who carried out the rewrite. A growing company needs permission to revisit decisions that once looked right. Early architecture is a working hypothesis. Protect it too fiercely and an organization can end up paying for its own loyalty every time it changes a feature.
The next introduction is to someone else’s system
By 2026, the company’s announcements increasingly involved fitting into systems other organizations already used. In May, Teton announced an integration with Nourish. Thorsen’s explanation included a short statement that travels well beyond either business: “Care teams need tools that work together, not in silos.”
It is an unromantic ambition for an AI founder, and a consequential one. Every extra interface asks someone to change a habit. Every separate login creates another small demand on attention. Connecting to an established platform means the new company accepts that it is entering an existing working day, with its own routines and loyalties.
In October, Teton announced another integration, this time with Inovonics’ EchoStream wireless network. Customers could connect existing hardware to Teton’s E-Call software. The network operated separately from building Wi-Fi. The announcement was dated October 14, 2026, a useful marker of where Thorsen’s company had arrived: working with installed infrastructure, rather than assuming customers would begin again.
These partnerships make the early loft feel less like a charming prelude and more like a recurring business problem at a different scale. First, build something. Then discover everything it must coexist with. The second task has a habit of taking longer.
A founder with room to change his mind
Teton’s own statement of values puts collaboration and mutual trust beside the demand to ship working products. That pairing suits the company’s public record: co-founders, engineers writing in their own names, returning investors and partners with their own systems to protect.
Thorsen’s story is most interesting where those relationships meet his willingness to question the machinery. The company has raised money, expanded its reach, rewritten software and opened part of its infrastructure. None of those steps removes the need to ask whether the next decision will make the product more useful.
Back in the shipyard loft, he could joke about refusing to become an ageing rock star stranded in an old ambition. Years later, that joke still has a job to do. It leaves room for effort without granting an idea immunity from criticism. A founder can take the work seriously and still allow the work to change.