Asad Tirmizi and the education of a robot

On the eve of leaving TikTok, Asad Tirmizi paused over his favorite view of San Francisco. He had a new company to start and a suspicion that quiet moments were about to become scarce. In a public post, he thanked his wife, his family, and his future co-founder, Lars Tingelstad. Then he described what he wanted to build: robots that could turn a person's instructions into useful physical work.

The scene had the familiar ingredients of a founder's departure announcement. The proposed destination was more particular. Tirmizi wanted machines to acquire skills through vision, touch, and experience, with language connecting those abilities to the person giving the instructions. He had been thinking about the problem for years. A resignation supplied the occasion to do something about it.

Today he is co-founder and chief executive of Trener Robotics, the company that began as T-ROBOTICS in 2024. Its software platform, Acteris, is aimed at industrial arms already doing, or waiting to do, factory work. His career makes that choice less surprising than it first appears. Long before conversational AI became a business proposition, he was studying what happens when a machine and a human need to understand the same physical event.

The fingertips came first

Tirmizi's education crossed three countries and three related disciplines. He studied aeronautical engineering at Pakistan's National University of Sciences and Technology, completing his undergraduate degree in 2008. A research engineering role at Air University followed. At Karlstad University in Sweden, he pursued a master's degree in controls from 2010 to 2012. Then came doctoral work at the University of Siena in Italy.

The sequence moves from building machines to governing their behavior to studying how their actions can be felt. Controls concerns the difference between what a system should do and what it actually does. Haptics brings touch into that exchange. In teleoperation, where a person operates a machine at a distance, the quality of that exchange helps determine how clearly the operator understands contact with the world.

His doctoral research examined different components of haptic feedback: vibration, sensations at the skin, and forces associated with movement. Tirmizi graduated in 2016, and his thesis received Italy's SIDRA doctoral award in systems and control. It is a specialist achievement with a practical subject. A machine can return information through the hand as well as through a screen.

That early work provides a useful way into his current interests. A robot loading a part has to deal with surfaces and contact, even when its user gives instructions in words. Touch remains part of the job. The vocabulary has expanded from academic papers to product demonstrations, but physical interaction still sits close to the center of Tirmizi's work.

A long apprenticeship, a short stopwatch

After Siena, Tirmizi led Factory-of-the-Future research at Flanders Make in Belgium. He later worked at Vicarious, the robotics company acquired by Google, and contributed to robotics and haptics at ByteDance. These were different settings for a related preoccupation: getting robot software to cope with the demands of physical tasks.

Tingelstad brought his own industrial background, including an associate professorship in robotic production at the Norwegian University of Science and Technology. Their collaboration reaches back to doctoral research. By the time they started a business together, neither was encountering the gap between a plausible robot movement and a dependable industrial operation for the first time.

Still, experience does not deliver a customer by itself. Tirmizi put an unusually precise number on that next step. In July 2024, he announced that the company had reached its first customer deployment in five months and 11 days, spending $38,000 along the way. Those figures were his account of the early effort, rather than a claim about what every future installation would cost.

The precision is revealing. He measured the journey in elapsed time and money spent, then put the satisfied user at the center of the announcement. He also described early backing and reaching the ABB challenge final. Those milestones mattered to the young business. Getting the software into somebody else's operation gave the work a different kind of consequence.

By November 2024, his public writing was drawing a firm distinction between demonstrating communication and executing a task reliably. Language models could help an operator express intent, he argued, while the robot's skill models had to carry the demands of industrial action. It was a useful piece of impatience from a founder selling a conversational interface: conversation still had to earn its place beside the machine.

What the sentence leaves out

Acteris gives that argument a product form. An operator describes a task; the platform uses deployable robot skills to configure and execute the work. Its tools include job sequencing, simulation, vision, and production monitoring. Trener currently lists compatibility with ABB, Universal Robots, and FANUC, giving its software a route into factories with different brands of equipment.

Consider the difference between asking for a batch of parts to be loaded and specifying each point an arm must visit. The first instruction describes a desired result. The second describes a movement plan. Tirmizi's business depends on putting enough knowledge between those two descriptions that an operator can work at the level of the task.

The distinction also explains why his background in touch matters. Language describes a part; contact supplies information about the act of handling it. A conversational system has to connect its description of the world to what a robot can actually perceive and do. A fluent sentence receives no special treatment from a metal fixture.

Trener's initial emphasis is CNC machine tending, the work around loading and unloading machine tools. In shops with frequent changes of part, variation arrives as ordinary business. A different order can require another setup. The economic question becomes how much engineering must be repeated before the arm can earn its keep again.

For Tirmizi, concentrating on that work is a deliberate choice. His team is training industrial skills with a specific use in mind. The platform's wider ambition can wait its turn while the first application meets the less romantic requirements of production: feasible motions, workable timing, and a process that operators can manage.

The dignity in a dull job

There is a human argument underneath his technical one. Tirmizi wants people to have more choice about repetitive industrial work. When he discusses automation, he returns to the person directing the machine and the knowledge that person brings. A robot's usefulness includes what it allows its operator to stop doing repeatedly.

He has also made room for humor in the account. Recalling Trener's machine tool innovation award, he described the team's amusement at winning recognition in machining while coming from robotics rather than the machinist's trade. The joke depends on the distinction: their work was to learn from task data and build a model, drawing on expertise that already existed in the shop.

His public manner mixes technical detail with a willingness to state a practical ambition plainly. He will discuss control frameworks and then return to whether somebody can give useful instructions in ordinary language. That movement between levels is part of the chief executive's assignment. The mathematician, the integrator, and the operator all have a stake in the same machine moving correctly.

“I like the robot to do the talking”

Asad Tirmizi, discussing future demonstrations

The line comes with a small joke of its own: his robots can literally talk. More substantially, it expresses his preference for showing a deployment. It is a sensible instinct for someone whose product must eventually leave the screen and share a workspace with other people's equipment.

Capital meets the workcell

The company has attracted money as well as attention. A $5.4 million seed round was announced in December 2024. In February 2026, Trener announced a $32 million Series A, co-led by Engine Ventures and IAG Capital Partners, bringing total funding above $38 million. Strategic participants included Cadence and Geodesic Capital through Nikon's NFocus Fund.

The planned uses were concrete: research through T-Labs, training additional skills, recruiting, and expanding markets and partners. The distinction between those activities matters. A stronger model still needs a route into a commissioned workcell. A broader sales network needs something repeatable to install. Tirmizi has to build both sides of that relationship.

Recognition has come from industrial settings. T-ROBOTICS was among the winners of ABB's 2024 AI Startup Challenge. Trener won the 2025 machine tool innovation award at EMO Hannover and the Industry Startup Forum award at Advanced Factories in Barcelona in 2026. These form a sequence of industry encounters as the company moves beyond its first customer.

More recent announcements extend the work toward computing infrastructure. In August 2026, Trener described its role as a founding AI Models and Software partner in AMD's Open Robotics Ecosystem, including skills running on Kria modules. In September, it joined Arm Total Design for Physical AI and work on a shared robotics capability framework.

For a founder trained in controls, the connection is straightforward: the computing beneath a model affects how quickly and predictably a machine can respond. A factory task has a timetable, and intelligence has to arrive in time to be useful. The hardware partnership belongs in the same story as the operator's sentence.

Back at the San Francisco view, the company was still an intention and the departure from TikTok an immediate fact. Tirmizi has since acquired customers, investors, awards, and a longer list of people who expect the next version to work. The ambition remains recognizable: let a person explain the job, and give the machine enough education to get on with it.