The long-horizon bet. AI applied to controls, sensing, and automation — the place software is heading whether industry is ready or not. Selective, exploratory, deliberately small.
The next decade of meaningful AI work happens where software touches the physical world — manufacturing lines, logistics, energy systems, agriculture, clinical environments. Generative models on a screen are easy now; closed-loop systems with sensing, control, and consequence are hard.
We don't run a robotics lab. We don't sell humanoids on TED stages. We co-build with operators who already understand a physical domain — and we bring the AI, software, and systems engineering they don't have time to staff for internally.
Most projects in this discipline never leave the operator's facility. That's the point.
Combining vision, depth, IMU, and process telemetry into closed-loop systems. From offline modelling to deployed controllers running at the edge.
Computer vision in industrial settings — defect detection, tracking, scene understanding. Not pretty demos; production-grade pipelines that survive a factory floor.
Building simulators that the real system actually transfers from. Reuses the spatial pipeline behind Witmodel and the data pipeline behind WMind.
Test rigs, calibration tooling, and the unsexy infrastructure that makes physical systems iterable on a software cadence.
We don't take engineering & robotics work as a vendor. We co-build with an operator partner who owns the domain, the facility, and the customer relationship — and we bring the engineering. Equity-for-build, capped scope, clean exits if the thesis doesn't hold.
If you're an operator with a physical problem, an existing customer base in a specific sector, and the willingness to co-build rather than buy consulting — we'd like to talk. Most weeks the answer will be "not now, not us" — but the right fit is rare and worth finding.