10,000+ hours a month: egocentric data collection across Southeast Asia
Client: Robot-learning data buyers (US)
We run egocentric and manipulation data collection across Malaysia, Indonesia, Vietnam, Thailand and the Philippines — more than 10,000 hours a month in commercial and household environments, on phones, egocentric glasses and tactile gloves, with capacity growing every month.
The problem
Robot-learning teams need large volumes of first-person data from real work — not staged demos. Collecting it is an operations problem: consented sites, trained participants, devices that are charged, clean and correctly mounted, quality control that catches duplicates and filler footage, and reliable upload. Most buyers have no people on the ground where this data can be gathered efficiently.
What we run
- Five countries: Malaysia, Indonesia, Vietnam, Thailand and the Philippines.
- More than 10,000 hours a month, with capacity growing month on month.
- Commercial and household environments: assembly lines, warehouses, kitchens, labs, retail and homes.
- Devices matched to the protocol: smartphones on head mounts for scale, egocentric glasses and multi-camera headsets for geometry and synchronised IMU, and tactile gloves for contact-rich manipulation.
- Consent first: employer and participant consent before the first session; sensitive areas, customers and screens excluded by protocol.
- Protocol QA: duplicate detection, filler and sensitive-content screening, session metadata, documented rejection reasons and fast participant retraining.
Why it works
The same team that runs collection also qualifies the hardware in Shenzhen. When a buyer asks for higher frame rates, stereo or synchronised IMU, we change the device, not just the instructions — and the next week’s hours reflect it.
Building hardware that has to learn? Talk to us.
Send a spec, a sketch, or just the problem. We reply within 24 hours (HKT), usually the same day.