Merit Robotics Let’s talk

THE DATA LAYER FOR PHYSICAL AI

Turn real work into robot-training data.

Real-world demonstrations ground our digital twins.

Our goal: scalable synthetic episodes for robot learning.

Discuss a pilot
Third-person view of a humanoid robot at a reconstructed industrial workcell in the Merit Robotics digital-twin prototype
DIGITAL TWIN / THIRD-PERSON VIEWActual frame from our workcell demo
Real work. Scalable intelligence.Explore the demo

01 / THE DEMO

From the factory floor
to a digital twin.

Real demonstrations anchor geometry and task context. Our prototype brings the workcell into a digital environment for comparison and iteration.

Real factory footage alongside robot-eye and third-person views of the reconstructed workcell.

WATCH THE WORKCELL DEMO

Real work.
Three perspectives.

Compare the real demonstration with the reconstructed robot view and the digital twin in motion.

35 seconds · No audio

Read a description of the demo

A worker handles rails and components at a factory workbench. The upper panels compare real head-camera footage with the reconstructed robot-eye view. The lower panel shows a humanoid robot carrying out the scripted task sequence in the digital workcell. This video demonstrates reconstruction and scripted kinematics; it does not establish physics or training-data validation.

Open video in a new view
First-person recording of a worker handling components at a real factory workbench
A

Real factory demonstration

The real-world task and workcell

Explore real-world demonstration data
First-person robot view showing tools and components in the reconstructed digital workcell
B

Reconstructed robot view

The task in a digital environment

PROTOTYPE TODAY

A real-to-digital-twin comparison using scripted kinematics. Physics validation and validated training episodes remain on the roadmap.

02 / OUR APPROACH

Grounded in reality.
Built toward scale.

A development workflow from real demonstrations to reusable workcells and, ultimately, validated robot-training data.

  1. 01

    Capture
    real demonstrations

    Start with real work, real geometry, and task context.

  2. 02

    Create
    the digital twin

    Use generative AI workflows to reconstruct the workcell.

  3. 03

    Calibrate
    with sensor data

    Bring the digital twin closer to its real-world reference.

  4. 04

    Generate
    task variants

    Reuse the digital workcell to explore new task variations.

  5. 05

    Validate
    training data

    Work toward validated episodes and recovery-data loops.

Our roadmap connects reconstruction, calibration, generation, and validation.

03 / WHY MERIT

Real context.
Reusable foundations.

Built for robotics and embodied AI teams developing a more capable physical AI future.

Grounded in real work

Real demonstrations anchor the geometry, objects, and task context behind each digital workcell.

Designed for reuse

Reusable workcells support new tasks and variants. Generative AI workflows aim to lower creation and data-generation costs.

Quality by design

Our approach puts rights, provenance, and quality checks at the center of data delivery.

BUILD WITH MERIT

Let’s start
with real work.

Explore a workcell pilot, discuss a data need,
or learn more about our approach.

Get in touch

Workcell setup + recurring data contracts
Roadmap: subscription + generation usage