Pakistani robotics startup World Context hosts first Narration Sprint in Lahore

Pakistani robotics startup World Context has hosted its first Narration Sprint in Lahore, where participants narrated footage collected from Pakistani factories to help create data for training robots to perform human tasks.

Participants worked with first-person footage recorded during World Context’s factory data-collection operations, describing workers’ actions and surroundings through audio narration.

The resulting data is intended for training vision-language-action (VLA) and vision-language models (VLMs).

Human narration adds context to recorded actions by identifying what a worker is doing, how the task is being performed and what is happening around them. This helps turn raw footage of human activity into structured data that can be used to train robots to understand and perform similar tasks.

The Narration Sprint was World Context’s second major event within a month, following its robotics hackathon at the University of California, Berkeley.

What did World Context say about the Narration Sprint?

“Don’t you see? We are bringing an entirely new class of employment home to Pakistan. Our people are training international artificial intelligence with their hands, their work, and now their voices,” Founder and Chief Executive Officer Sherdil Ali Khan said.

“The event was a success and we hired multiple participants as they brought expertise and energy. Pakistanis, through data production, are nurturing and guiding robots through their infancy.”

“Why aren’t more people talking about this? Why aren’t more people doing this? Data is a trillion dollar opportunity for the country,” he added.

“I am proud to be a pioneer of the field for us. Many to come, God willing.”

World Context collects footage of factory workers performing tasks across different sectors and sells the data to U.S. robotics and AI companies.

The company aims to expand technology exports between Pakistan and the United States.