About
We build the data layer for Physical AI
Khenda captures and structures the manipulation data humanoids learn from, inside the plants where that work actually happens.
Our story
Khenda was founded in 2021 around a single observation: the largest untapped source of real manipulation data in the world is not a lab and not a warehouse of teleoperation rigs. It is the shop floor, where hundreds of millions of people perform precise physical work every shift.
We started inside factories, building a process analysis product that manufacturers actually pay for and run in production. That gave us something a data company almost never has: standing inside real plants, a working relationship with the people on the line, and a legal framework for recording them that their employers and their works councils will sign.
As humanoid programmes ran into the data wall, it became clear that the position we had built was the scarce asset. Raw egocentric video is now commoditised. Task-directed capture inside live production, labelled at the action level, with legally clean consent, at scale, is not.
Mission
To make real human work legible to machines, so that a robot can learn a physical skill from the people who already do it well, and so that the people who taught it are credited, consenting and paid.
What makes Khenda different
We are already inside the plants
Our commercial platform runs inside production manufacturing plants today, across seven countries. Access is not a business development problem we still have to solve.
A contracted factory network
A large and growing network of production factories that are contracted and ready to share data, spanning multiple sectors and hundreds of distinct work stations and process types.
Consent built to be reviewed
Operators consent explicitly and individually and are compensated. Employers sign data agreements. The framework was developed with labour law counsel to be KVKK and GDPR compatible.
Granted patent
Granted U.S. Patent No. 12,450,905 B2, covering periodic and temporal action segmentation.
Certified operations
SOC 2 Type II and ISO 27001, with GDPR and KVKK compatible processes.
No special hardware on the line
Capture works with standard cameras. No depth sensors and no motion capture suits are required for the data we deliver.
Proof points
- Founded
- 2021
- Patent
- Granted U.S. Patent No. 12,450,905 B2, covering periodic and temporal action segmentation.
- Production footprint
- Commercial deployments inside production manufacturing plants across seven countries.
- Sectors
- Automotive, electronics, parts manufacturing, logistics, home appliances, heavy industry.
- Certifications
- SOC 2 Type II, ISO 27001, with GDPR and KVKK compatible processes.
- Offices
- Ann Arbor, Michigan and Istanbul.
For manufacturers
Alongside the data business, our commercially deployed process analysis platform for manufacturers is a product in its own right, used inside production plants to improve how work is done.
Visit khenda.com(opens in a new tab)Offices
Ann Arbor
Michigan, USA
Istanbul
Türkiye
Work with us
Whether you are training a humanoid, running a plant, or researching robot learning, the conversation starts the same way: what task, and what does the data have to look like.
Get in Touch