Finishing a wall is difficult for human workers and even harder for robots. Taping, mudding, and sanding are slow, exacting work, and the skilled trades who do it are in short supply. It is also the work that determines how a finished space looks and holds up, which is why it has always depended on experienced hands and long hours. Canvas Robotics is changing that.
The Work Canvas Is Automating
Canvas, a JLG company, is bringing automation to drywall on construction sites. Their robots handle drywall finishing on real job sites, taking on the taping, mudding, and sanding that crews have always done by hand. This is not a demo in a controlled space. It is finishing work on active construction sites, held to a quality standard that professional crews can stand behind.
That combination, real sites and real finish quality, is what makes the problem hard.
Why the Job Site Is a Harsh Environment
A job site is never the same twice. Surfaces vary from one wall to the next. Layouts change between rooms. Lighting shifts through the day. Conditions drift in ways no fixed program can fully anticipate.
Automating this work means building robots that hold up when conditions change, not robots that perform once in a controlled setting. A finishing robot has to read the surface in front of it, reason about how its tools and materials will behave on that surface, and adjust as the environment moves around it. Taping, mudding, and sanding are contact-rich tasks where force, geometry, and material response all matter, and all of them vary from one wall to the next.
Where TorqueAGI Fits
This is where we come in. TorqueAGI contributes the Physical AI infrastructure that helps robots like Canvas reason through that variability and perform reliably in the conditions real sites demand.
Instead of memorizing what environments look like in training, the TorqueField model reasons about geometry, surfaces, and the forces involved to make efficient use of data in training the model for the robots.
Building Together
The shortage of skilled labor is not going away, and the work itself is not getting easier. Robots that can do it reliably, on real sites and in real conditions, give construction teams a practical way forward. We are glad to be building that alongside the Canvas and JLG teams.
If you are working on robotics for construction or other demanding real-world environments, we would love to connect. Feel free to reach out to us directly.
