News

Machinery and Technology

Indian Garment Workers Are Teaching Robots To Stitch

• By  Textile Excellence
• October 31, 2026

From garment workers filming their stitching techniques to barbers recording their hand movements, a new gig economy is emerging around training physical AI. It pays workers today, but raises some uncomfortable questions about tomorrow.

A garment worker stitches fabric every day. A barber handles a razor with practised precision. Nothing unusual there. Except that some of them are now doing their jobs with a camera strapped to their heads - not to make social media videos, but to teach robots how to do the same work.

Welcome to a new and rather unexpected corner of India's gig economy: recording human activity to train artificial intelligence.

The demand comes from companies developing physical AI - systems designed to control robots and perform tasks in the real world. Unlike conventional AI, which learns from text, images and digital information, these systems need to understand how human hands move, how much force to apply and how people interact with objects.

And where better to collect such information than from people who have spent years mastering practical skills?

After a decade in the textile industry, stitching and cutting fabric have become second nature to garment worker Mohammed Arman. Now, he records these familiar movements for AI training. He sees it as a practical way to earn a little more without taking on an entirely different job.

A new market for human movements

The footage is collected by companies such as Bengaluru-based Rashdan Technologies and supplied to technology firms and research organisations. The tasks are not limited to textile work. Data collectors are being sought for activities ranging from embroidery and pottery to agriculture, household chores and small-business operations.

Rashdan Technologies pays around $5–10 for eight to ten hours of recorded work, depending on the assignment. The company says it takes steps to protect workers' identities, including recording hands rather than faces and blurring faces where necessary.

The demand is growing as robotics developers seek real-world data to train machines. A robot cannot learn to stitch a garment, grip a tool or handle delicate material simply by reading instructions. It needs examples of how people perform these tasks, including the movements, coordination and physical judgement involved.

That is why ordinary workplaces are becoming valuable training grounds. The knowledge held in a worker's hands, often acquired through years of repetition, is being converted into data that machines can learn from.

Humyn Labs, another company in this space, collects and processes human-generated data across countries in Latin America and Southeast Asia. It filters large volumes of footage, identifies useful material, labels it and prepares it for clients. The company says it has 96,000 data collectors across countries.

This points to a much larger opportunity than simply wearing a camera. The value lies in turning raw recordings into reliable, structured training data. As competition increases, India could move beyond supplying footage and develop capabilities in data processing, validation and quality standards.

The uncomfortable question

There is, however, a bigger issue lurking behind this new source of income. Are workers being paid to train the very machines that could eventually replace them?

It is too early to predict how quickly physical AI will change employment. Robots may take over some tasks, assist workers with others or create entirely new jobs. The outcome will vary across industries and applications.

But the possibility cannot be brushed aside. A worker may see the camera as an easy way to earn extra money today, without necessarily knowing how the recorded movements could be used years later. Once the data has been collected and incorporated into training systems, the worker may have little control over its future use.

Consent, privacy, compensation and ownership of the data therefore need serious attention. The recording process may begin with a camera focused only on hands, but some companies are already asking for multiple cameras to capture more of the body and surrounding environment. That raises fresh questions about how much of a person's work and workplace should be recorded.

For India's textile and garment workforce, the development is particularly interesting. The country has a vast pool of skilled workers whose practical knowledge could become an asset in the emerging robotics economy. But remaining merely a source of low-cost human footage would be a missed opportunity.

The bigger prize lies in building the capabilities to process, organise and validate that data, and eventually participate in setting the standards for how it is used.

For now, workers are earning a few extra dollars by showing machines how to do what humans do best. The real test will be whether this new economy creates better opportunities for those workers, or simply makes their skills easier for machines to copy.

The bigger prize lies in building the capabilities to process, organise and validate that data, and eventually participate in setting the standards for how it is used. For now, workers are earning a few extra dollars by showing machines how to do what humans do best. The real test will be whether this new economy creates better opportunities for those workers, or simply makes their skills easier for machines to copy.