Embodied AI

Klerix

On-board intelligence for industrial and warehouse robots.

Overview

What Klerix does.

Klerix puts a compute brain on the machine. Perception, reasoning and safe control run on a ruggedized module bolted to the robot itself, so a warehouse or factory arm can handle a shifted pallet, a new SKU or a change in lighting without an engineer reprogramming the cell. Skills are trained in simulation first, validated against safety rules, then deployed to the fleet.

The platform ships as five products: Cortex, the on-robot runtime; Aegis, an independent safety monitor that records a replayable trace of every motion; Mirror, the simulation and skill studio; Helm, fleet analytics and teleoperation; and Anchor, per-robot identity and signed over-the-air updates.

At a glance
DetailValue
StageSeed
CategoryEmbodied AI
Year2026
Raised$2M
ModelOutcome pricing per task completed, entered through a 60-day paid pilot on one work cell.
Measured byTasks completed per shift, unsafe-motion incidents, fleet uptime, cells expanded per customer.
An inspection station on a production line
The buyer Industrial operators in Europe and the Gulf, paying in hard currency
Close view of a compute module
The wedge One work cell, one paid pilot, sixty days
Why we invested

The thesis behind Klerix.

The physical work that moves goods through a warehouse still runs on two bad options. Fixed automation is fast and rigid: a caged cell pinned to one part and one layout, where any change in SKU means weeks of re-integration. Manual labour is flexible and scarce, priced at $18 to $30 an hour with turnover above 60 percent. Neither option absorbs a floor that changes weekly.

Klerix sells the missing middle. The robot keeps its flexibility because the intelligence sits on the machine, and the operator keeps their audit trail because a separate safety monitor checks every motion before it executes. Pricing follows throughput rather than seats, so the customer pays for tasks completed instead of licences purchased.

This fits our test cleanly. The buyer is a European or Gulf industrial operator paying in hard currency, so the revenue does not depend on the local market. The build advantage is real: an engineering team fluent in the NVIDIA robotics stack, from Isaac Sim training through TensorRT optimisation to Jetson deployment, staffed at Pakistani cost against American and German competitors. The wedge is narrow enough to close, since the first sale is one work cell rather than a whole floor. And every deployment feeds a dataset of real interaction traces that the next model trains on.

Containers stacked at a logistics terminal
Five products One simulation to edge to fleet loop
Team

The people building it.

Founded in 2025 by an AI engineer and a robotics engineer who met over the same problem from opposite sides: the models were good enough, but nothing carried them safely onto a machine that moves. The team builds from Pakistan and deploys into food, manufacturing, 3PL and automotive floors.

Klerix is a member of NVIDIA Inception and builds natively on the stack, using Isaac Sim and Omniverse for digital twins, Isaac ROS for perception and localisation, TensorRT for edge inference, and Jetson Orin and Thor class modules on the robot.

  • Website klerix.com
  • Industry Embodied AI
  • Round Seed
  • Raised $2M
  • Year 2026
  • Founded 2025
An engineer reviewing sensor output on screens
Built in Pakistan Deployed on floors abroad

Building something similar?

Building in robotics or embodied AI? We want to hear about it.