Anjali Sardana, the founder of instant home-services startup Pronto, has set up a new venture, Provision Labs, to work on one of the hardest problems in artificial intelligence: getting enough good real-world data to train machines that act in the physical world. The San Francisco-based company will focus on the data, evaluation and deployment layer for physical AI, and will work with developers building AI models for robots and other real-world applications.

The venture was first reported by The Economic Times in late October. Sardana will be chief executive of both Pronto and Provision Labs.

Backed by Pronto’s investors

Provision Labs’ website lists Lachy Groom, General Catalyst, Bain Capital Ventures and Glade Brook Capital among its backers. All four have also invested in Pronto. The size of the funding and other financial terms have not been disclosed.

Why data is the bottleneck for physical AI

Large language models learned from the vast amount of text already on the internet. Robots and other physical AI systems have no such ready-made library. To learn how to fold laundry, wash utensils or move objects around a home or factory, they need large amounts of recorded real-world activity, captured across many environments and tasks, along with reliable ways to test whether a trained system can actually do the job.

Provision Labs says its edge comes from running complex, scaled physical-world service operations. In its job listings, the company says it wants to capture diverse, “in the wild” pre-training data for robotics and to cover the rest of the stack where real-world complexity matters, including physical-world evaluations, multimodal data collection, post-training and deployments. Its plans include in-house data-capture hardware, human and automated annotation and quality-control pipelines, and models for tasks such as 3D pose reconstruction and SLAM (simultaneous localisation and mapping).

The Pronto connection

Pronto, founded in 2025, connects households with trained and background-verified professionals for cleaning, laundry, utensil washing, meal preparation, car washing and gardening. In July 2026, the company confirmed it was expanding Pronto Verified, an opt-in service in which professionals wear cameras that record service sessions, with anonymised footage used to train AI models.

That kind of first-person footage of everyday household work is exactly the type of data physical AI developers are looking for, which makes the new venture a natural extension of Pronto’s operations.

An India link to a global race

Although Provision Labs is headquartered in San Francisco, it is also hiring machine learning engineers in Bengaluru to design new approaches to robotics data, learning, perception, evaluation and control. The move adds to a growing list of India-linked bets on physical AI, as global developers race to give robots the same kind of data advantage that powered the rise of generative AI.

For enterprises and policymakers watching the space, the launch is a reminder that the next phase of AI competition may be decided less by model size and more by who controls high-quality, consented real-world data, and how responsibly it is collected.

Be a part of Elets Collaborative Initiatives. Join Us for Upcoming Events and explore business opportunities. Like us on Facebook , connect with us on LinkedIn and follow us on Twitter.

"Exciting news! Elets technomedia is now on WhatsApp Channels Subscribe today by clicking the link and stay updated with the latest insights!" Click here!