CarbonSix  ·  Physical AI for Manufacturing  ·  SigmaKit

Bringing Six-Sigma robot dexterity and intelligence to manufacturing automation.

We capture operator expertise with AI-powered imitation learning and turn it into robot skills that survive the realities of the factory floor in high-variance manufacturing environments.

SCROLL
01
About us

Demand for automation is enormous — yet much remains unautomated.

Conventional robots repeat a fixed path. Shift the part, the model, or the condition and they stop — which is why High-Mix, High-Variance work is still done by hand.

What traditional automation solves
  • Predictable object motion
  • Clean, aligned environments
  • No product changeover
  • Motions that are easy to design

High predictability · Low variance

Demand for automation
Frontier — still done by hand

High-variance work

Wiring, film, high-mix, unstructured — the majority of the demand.

Traditional automation ✓

Automatable today

CarbonSix enables fast teaching of high-variance manufacturing tasks in the language of data.

On the floor · not in a demo lab
Robots that learn the way
technicians teach.
The factory already holds the knowledge. SigmaKit turns operator know-how into deployable robot skill — without code, AI expertise, or special equipment.
02
The product · SigmaKit

One kit: teach, train, deploy.

SigmaKit is delivered as a complete package — combining AI software with robotic hands, teaching devices, and sensor modules — so it can be used easily without deep expertise in robotics or AI.

Demonstrate the task, review the data, train a Skill, and deploy it to the line. Depending on the task, a model can be generated in less than a day.

No complex coding. No AI expertise. Accessible for anyone on the shop floor.

SigmaKit in detail
SigmaKit components: AI software, robotic hands, teaching devices and sensor modules
The complete SigmaKit package — AI software, robotic hands, teaching devices, sensor modules.
Applications

Hard tasks, proven on real lines.

Deformable cables, thin films, reflective hooks, moving conveyors — the tasks conventional automation gave up on.

03
The team

Built by people who've shipped.

MIT, Yale, Caltech, SNU, KAIST — SUALAB, Cognex. Founders who have taken robotics and AI from research to exits and global deployments, advised by Russ Tedrake and Clément Gosselin.

Newsroom

Latest from CarbonSix

Get in touch

Let's automate the
un-automatable.

Tell us about your hardest station. We'll tell you whether a robot can learn it — and how fast.