CARPL.aiThe Story
CARPL.ai has raised $10 million in a Series A round led by the International Finance Corporation, the private-sector arm of the World Bank Group, with participation from existing investor Stellaris Venture Partners and other investors who were not named. The round was announced on 23 July 2026. The company said the money will go towards strengthening its technology stack, accelerating product development and expanding its global sales and delivery teams. The valuation, the dilution and the instrument have not been disclosed. Neither has revenue, nor the commercial terms under which hospitals or AI vendors pay to use the platform. The company had previously raised $6 million from Stellaris Venture Partners and angel investors, which takes disclosed funding to $16 million. Vidur Mahajan founded the company in 2018, though its own 2024 announcement described it as operational since 2021. CARPL was incubated at Mahajan Imaging and Labs, a large Indian radiology services provider, as its technology division, under the guidance of Harsh Mahajan, the founder's father. Corporate records list the legal entity as CARPL.ai Inc., registered in San Francisco. The platform allows hospitals and radiologists to discover, deploy, manage and monitor medical imaging AI applications from multiple vendors through a single interface. Its marketplace now hosts more than 300 applications from over 100 vendors, against more than 50 developers and 100 applications reported in February 2024. CARPL.ai says it serves 30 hospitals and medical centres across North America, Europe, the Middle East, India, Southeast Asia, Australia and Latin America. Named customers include the Singapore government, RadNet, I-MED Radiology, Fleury, Medica and the Clinton Health Access Initiative in India.
Why It Matters
CARPL does not sell artificial intelligence. It sells the removal of an integration cost, and that cost is a large part of why radiology AI has been slow to reach hospitals. Consider what a vendor with one good algorithm faces. To sell it into a hospital it has to connect to that hospital's picture archiving system, pass an IT security review, produce local validation data, clear procurement, and train the radiologists who will use it. That work can cost more than the licence itself, and it has to be repeated at every hospital. From the hospital's side the same problem runs in reverse. Each additional algorithm is another integration to build and another vendor to monitor for drift. A single connection carrying 300 applications collapses that cost for both parties. The mechanism is real, and it explains why large providers have signed up. What has not been disclosed is who pays. If hospitals pay a platform fee, revenue scales with 30 customers. If vendors pay for distribution, it scales with 100 vendors competing for position on the same shelf. Those are different businesses, with different margins and different bargaining power, and the announcement does not say which one this is. Nor does it say what the platform takes. Without a take rate, 300 applications and 30 hospitals are counts rather than economics.
The Strategic Read
The market assumption being underwritten is that medical imaging AI stays fragmented. Aggregation is worth something in proportion to the mess it resolves. In early 2024 there were roughly 200 companies building radiology AI and several hundred cleared products, and CARPL's argument was that no hospital could evaluate that field alone. The argument was correct. The supply side has since tripled on CARPL's own shelf, from around 100 applications to more than 300. The demand side has not moved at the same rate. Thirty hospitals and medical centres, spread across seven regions, five years after the platform became operational. Each is a serious institution and several are very large, so revenue per account may well be substantial. But a marketplace with a hundred sellers and thirty buyers is not yet functioning as a marketplace. It is an enterprise software company with a large catalogue, and the vendors on that shelf are the ones absorbing the imbalance. Where value is captured depends on the picture archiving relationship, and so does the risk. Mahajan describes the platform as deeply integrated with leading enterprise PACS providers. Those integrations are difficult to build and they are the closest thing here to a moat. They are also a dependency. Sectra, Philips, GE, Siemens and Microsoft's imaging network can each assemble a catalogue of the same third-party algorithms and bundle it into software that hospitals already run. A partner that controls the workflow can become a competitor without buying anything. The second risk is consolidation in the models themselves. CARPL's value rises with the number of narrow point solutions a hospital wants to operate. General-purpose imaging models that read many findings in one pass would reduce that number, and a layer built to manage fragmentation thins as the fragmentation does. IFC leads this round with an explicit emerging-markets diagnostics mandate, which points growth towards the customers least able to pay for it.
For daily, sharp analysis of the biggest moves in the Indian business and startup ecosystem, follow StartupFox.
