Provision Labs, the physical artificial intelligence (AI) venture founded by Pronto founder Anjali Sardana, has acquired Mumbai-based hardware startup Basilisk Industries for an undisclosed amount. The acquisition brings Basilisk founder Shourya Watwani and his team into Provision, strengthening the company’s capabilities in perception technology, hardware development and tactile data collection.
The deal comes as Provision builds infrastructure to collect real-world data for training and testing AI systems designed to perform physical tasks through robots and machines. Basilisk’s work includes a 3D hand-tracking model, redesigned perception hardware and an early version of a lightweight tactile glove that captures touch-related information. The acquisition gives Provision additional engineering capabilities as it develops its physical AI business.
Provision Labs Acquires Basilisk Industries
Anjali Sardana announced the acquisition in a LinkedIn post on November 2, 2026. The financial terms of the transaction have not been disclosed.
Basilisk founder Shourya Watwani will lead Provision’s hardware efforts and work on the company’s in-house post-processing models. The Basilisk team is also joining Provision as part of the transaction.
Founded in 2023, Basilisk Industries focused on perception technologies that help machines understand and interpret their surroundings. Its work is relevant to robotics and physical AI applications, where machines need to recognise objects, track human movements and interact with the physical environment.
According to Sardana, the Basilisk team has already contributed to several projects at Provision, including:
- Developing a 3D hand-tracking model.
- Redesigning Provision’s perception hardware.
- Building the first version of a lightweight tactile glove to capture touch-related data.
- Strengthening the company’s hardware and perception engineering capabilities.
The acquisition brings these capabilities under Provision’s operations rather than requiring it to develop every component independently.
What Is Provision Labs Building?
Provision Labs is a San Francisco-based physical AI venture established by Sardana, who also founded Indian instant home-services startup Pronto.
The company focuses on collecting real-world data that can be used to train and evaluate AI systems designed to operate in physical environments. Unlike conventional AI systems that primarily process text, images or digital information, physical AI systems need to understand how people and objects interact in the real world.
For example, a robot designed to fold clothes needs information about how a person picks up fabric, adjusts its position, applies pressure and completes the folding movement. Similar data is required for tasks such as cleaning, handling objects, preparing food and operating equipment.
Provision plans to source training data from professionals working across different sectors globally. It aims to work with developers building physical AI models and provide data that can help these systems perform real-world tasks.
The acquisition of Basilisk adds a hardware development component to this strategy. It potentially allows Provision to build specialised devices for collecting different types of physical interaction data.
Why the Tactile Glove Matters for Physical AI
One of the most important developments associated with the acquisition is Basilisk’s lightweight tactile glove.
Most computer vision systems rely on cameras to understand the environment. Cameras can capture the position, movement and appearance of objects, but they cannot directly measure the physical sensation of touch.
A tactile glove can help capture information about how a person interacts with objects, including contact and touch-related signals. This type of information could help developers train robots to handle objects more accurately.
Consider a robot learning to pick up a glass. Visual information can help it identify the glass and estimate its position. However, information about contact and applied pressure can help the robot understand how firmly it needs to grip the object without breaking it.
The same principle applies to handling clothes, kitchen utensils, tools and other everyday objects.
| Technology | Information captured | Potential use in robotics |
|---|---|---|
| Computer vision | Images, object positions and movement | Recognising objects and understanding environments |
| 3D hand tracking | Hand movement and spatial positioning | Learning human hand movements and manipulation |
| Tactile sensing | Touch and contact-related information | Improving object handling and grip control |
| Perception hardware | Data from physical sensors | Helping machines interpret their surroundings |
Combining visual, movement and touch-related information can provide developers with a more complete understanding of how humans perform physical tasks.
However, the commercial performance of Provision’s hardware and the accuracy of its data systems have not yet been established publicly.
Provision Expands Hiring Across San Francisco, Toronto and Bengaluru
Alongside the acquisition, Provision is expanding its research and engineering teams across San Francisco, Toronto and Bengaluru.
The company is hiring researchers and engineers in several areas, including:
- Computer vision
- Perception technology
- Tactile data collection
- Environment reconstruction
- Hardware engineering
These hiring areas indicate that Provision is investing in multiple layers of physical AI infrastructure, from collecting information in the real world to processing it for model training and evaluation.
The addition of Basilisk’s hardware team could help accelerate this development by bringing existing engineering expertise and prototypes into the company.
Pronto’s Household Recording Pilot Raises Data Privacy Questions
Provision’s development comes against the background of scrutiny surrounding a household recording pilot conducted by Pronto.
The home-services startup previously introduced an opt-in initiative in which service professionals wore cameras while performing household tasks such as cleaning, washing dishes and folding laundry. The recordings were intended to help generate real-world data for physical AI and robotics training.
The initiative raised questions about customer privacy, worker consent, data storage and the commercial use of recordings captured inside private homes.
Pronto said the pilot was voluntary, available only to customers who actively opted in while booking, and covered less than 0.01% of its users at the time. The company also said the recordings were anonymised and would not be shared with third-party AI laboratories.
The development of tactile hardware introduces another dimension to the broader data-collection discussion. As companies collect more detailed information about how people perform physical tasks, they will need clear processes for obtaining consent, protecting personal information and explaining how collected data will be used.
There is no public confirmation that Provision Labs will use Pronto’s household recordings as part of its own datasets or that the two businesses have a formal data-sharing arrangement. Their operations and data practices should therefore be considered separately unless further details are disclosed.
Physical AI Is Creating a New Market for Real-World Data
The acquisition reflects a growing interest in physical AI, a field focused on developing AI systems that can interact with the physical world.
Robotics developers need large volumes of real-world information to train systems that can navigate environments, manipulate objects and perform tasks reliably. Simulated environments can help developers test robotic systems, but real-world data remains important for understanding the complexity of everyday activities.
This is creating opportunities for companies that can collect, organise and process physical interaction data.
Provision is positioning itself in this emerging market by combining data collection with hardware development. Basilisk’s perception technology and tactile glove could become part of the infrastructure used to capture information for robotics developers.
However, the business remains at an early stage. Provision has not publicly disclosed the acquisition value, its commercial customer base or the revenue generated from its data services. The acquisition should therefore be viewed as a capability-building move rather than evidence of an established large-scale business.
The Bigger Picture
The Provision-Basilisk acquisition highlights a shift in AI development from digital information towards data generated through physical interactions. As companies build robots capable of performing household, industrial and commercial tasks, they need more than images and text. They also need information about movement, contact, pressure and how humans interact with objects.
For India, this creates an opportunity to participate in the emerging physical AI supply chain through hardware development, data collection and engineering services. But the sector will also need to address questions around privacy, worker compensation, data ownership and the distribution of value created from human-generated training data.
Looking Ahead
Provision’s next challenge will be turning its hardware and data-collection capabilities into a reliable offering for robotics developers. The integration of Basilisk’s team, further development of its tactile glove and expansion of its engineering workforce will be important steps in that process. The company has not yet announced a detailed commercial roadmap or disclosed the financial terms of the acquisition.
The broader opportunity will depend on whether physical AI developers are willing to pay for high-quality real-world datasets and whether Provision can collect them at scale while maintaining strong data protection practices. The acquisition gives the company additional hardware expertise, but its long-term position in the physical AI market will depend on execution, customer adoption and the quality of the data it can deliver.
Get the day’s top stories in your inbox
One concise email. No spam, unsubscribe anytime.
