Startup Deep Dive : CARPL.ai — how a Delhi company became the app store for radiology AI

More than 700 radiology AI tools have cleared the United States FDA, and yet in most hospitals not one of them touches a live scan. CARPL.ai, a New Delhi company that writes none of those algorithms itself, has turned that gap into a business: in July 2026 it raised a $10 million (about ₹96 crore) Series A led by the World Bank’s investment arm, the International Finance Corporation, taking its disclosed funding to $16 million.

The pitch is almost boring in its simplicity. Radiology AI is everywhere on paper and almost nowhere in the scan room, because every algorithm speaks a different technical language and every hospital’s imaging plumbing is its own tangle. CARPL sells the layer in between, a single platform through which a hospital can discover, test on its own data, deploy and monitor more than 300 AI applications from over 100 vendors. This is the story of how a family radiology business in Delhi turned into a piece of global healthcare infrastructure, what the public record actually shows about its size, and where the model could still break.

Quick facts

What CARPL.ai actually does

CARPL.ai runs a vendor-neutral platform that lets hospitals, imaging chains and radiology groups find, validate, deploy and monitor artificial-intelligence tools for medical imaging through a single interface. Think of it as an operating system and app store sitting between the hospital’s scanners and the many companies that build AI to read those scans. As of the July 2026 fundraise announcement, the company said the platform carried more than 300 radiology AI applications from over 100 AI partners. The problem it targets is documented across the sector:

• Over 700 radiology AI tools already have US FDA clearance, yet adoption in day-to-day clinical work remains low (TechCrunch, February 2024).
• Every algorithm has its own integration, data format and monitoring needs, so hospitals struggle to run more than a handful at once.
• Medical imaging generates over 80% of healthcare data, per CARPL’s own July 2026 materials, making it the densest place to apply AI, and the hardest to manage.
• CARPL’s answer is a “DEV-D” workflow, discover, explore, validate and deploy, plus a universal AI viewer that lets radiologists interact with algorithm outputs inside one screen (HITConsultant, February 2024).

The origin: from Mahajan Imaging to a marketplace

CARPL did not begin as a startup pitch; it began as an internal tool. Its name is a nod to the CARING (Center for Advanced Research in Imaging, Neurosciences and Genomics) Analytics Platform, the software used inside the founder’s family radiology business to validate and deploy AI on real patient scans. Dr Vidur Mahajan, a physician with an MBA from the Wharton School and more than a decade in diagnostics, ran Mahajan Imaging, a Delhi radiology chain, where he grew revenue tenfold to a company-stated $20 million over ten years, according to TechCrunch and HITConsultant.

The founding insight came from watching that AI adoption fail from the inside. Dr Mahajan and his father, the radiologist Dr Harsh Mahajan, kept trying to put algorithms into live clinical use and kept hitting the same wall: the tools worked in demos but were painful to test, integrate and trust at scale. The realisation was that the bottleneck was not the AI itself but everything around it, and that this middle layer could be a product other hospitals would pay for. CARPL was spun out to sell that layer to the world rather than keep it in one imaging chain. The platform launched publicly in 2021, with the legal entity dating to 2018.

The hard part: selling software nobody had a budget for

The gap between “obviously useful” and “someone will pay for it” is where most health-tech companies die, and CARPL’s early years sat squarely in that gap. Two structural problems shaped the difficult stretch, both visible in the public record.

First, there was no established line item for what CARPL sells. Hospitals buy scanners, PACS systems and reporting software; a marketplace-and-monitoring layer for third-party AI was a category that barely existed when the company launched in 2021. Selling it meant educating buyers on a problem many did not yet feel acutely, in a market where, as TechCrunch noted in February 2024, more than 200 companies were building radiology AI and over 700 tools had FDA clearance but real-world use stayed low. CARPL was, in effect, betting that the flood of algorithms would force hospitals to want a control layer before that demand fully arrived.

Second, the numbers stayed small. Tracxn’s filing-based data shows operating revenue of just ₹10.9 crore for the year ended 31 March 2024, roughly a decade after the underlying software work began and three years after the public launch. Tofler’s summary of the same year shows revenue slipping about 2.9% and the bottom line deteriorating sharply, with net profit down 201% and EBITDA down 194% against the prior year. For a company positioning itself as global infrastructure, that is a reminder that the enterprise healthcare sales cycle is long, and that scale in “number of AI apps” had not yet translated into scale in rupees.

The turning point: the IFC round

The clearest inflection in CARPL’s public story is the July 2026 Series A. In that round the company raised $10 million, about ₹96 crore, led by the International Finance Corporation, the private-sector arm of the World Bank Group, with existing backer Stellaris Venture Partners participating. Two things make it a genuine turning point rather than just a bigger cheque.

On one side of the event sits a company that, as recently as February 2024, described itself as post-revenue with roughly 100 apps from about 50 developers and a seed round of $6 million. On the other side, by July 2026, sits a platform stating more than 300 applications from over 100 partners, used by four of the world’s five largest private radiology groups, with government deployments across Brazil, India, Singapore, Spain and the UAE. The IFC’s involvement matters beyond the money: a World Bank institution backing an Indian deep-tech company signals that the buyers CARPL wants, public health systems and large enterprises in emerging and developed markets, are now in reach. The round roughly tripled the app count on the platform versus two years earlier and doubled disclosed funding to $16 million.

The money behind it

• Seed, February 2024: $6 million (about ₹57.6 crore), led by Stellaris Venture Partners with angel investors drawn from Novo Holdings, LeapFrog, Bain & Company, Boston Consulting Group and UnitedHealth Group, per TechCrunch (February 2024). The company was described as post-revenue at this stage.
• Series A, July 2026: $10 million (about ₹96 crore), led by the International Finance Corporation, with Stellaris Venture Partners participating (YourStory, Medical Dialogues, Inc42, July 2026).
• Total disclosed funding: $16 million (about ₹153.6 crore) as of July 2026, per Crunchbase and multiple round reports.
• Valuation: not disclosed. Reports noted the valuation, dilution and instrument were not made public.
• Recognition: CARPL was part of the YourStory Tech30 cohort of 2022.

What each backer changed is worth naming. Stellaris seeded the company and stayed in through the Series A, providing continuity. The IFC brought both capital and the institutional credibility of a World Bank body, aligned with CARPL’s stated push into public health systems and emerging markets.

How it makes money

CARPL is a marketplace, so its economics sit on top of other companies’ products rather than its own algorithms. The public description of the model breaks down like this:

• Platform access and validation: customers pay to run third-party AI algorithms on their own data inside a secure environment behind their own firewall, per TechCrunch (February 2024). This lets a hospital test tools before trusting them, and it is a recurring, software-style relationship.
• Clinical deployment and monitoring: once an algorithm is chosen, CARPL monetises the ongoing deployment, integration and monitoring in live workflows, the part hospitals find hardest to do alone.
• Vendor-neutral positioning: because CARPL builds none of the AI itself, it can host competing tools side by side, which is its core pitch to both hospitals (choice) and vendors (distribution).
• The part people get wrong: the headline metric of “300+ apps” is a supply figure, not revenue. A marketplace only earns when hospitals actually run and pay for tools at volume; app count and revenue can move independently, which the FY24 figures show clearly.

The numbers

CARPL is a private, early-stage deep-tech company, and its granular multi-year financials are not fully public; some line items sit behind paid filing databases. The figures below are limited to what could be verified from filing-based sources this month, in ₹ crore, with the year each covers.

The picture the data supports is consistent: revenue was roughly flat and modest in FY24 while losses widened, which is normal for a company spending ahead of a long enterprise sales cycle, but it also means the ₹96 crore Series A is funding a business that had not yet found revenue scale as of its last filed year. Where FY25 and FY23 absolute figures could not be verified from open sources, they have been left out rather than estimated.

Where the money comes from

The revealing split for CARPL is geographic and by customer type rather than by product line, and it runs against the usual assumption about an Indian startup.

• Global, not domestic-first: the platform is used across North America, Latin America, Europe, Asia-Pacific and emerging markets, with named clients spanning the US, Australia, Brazil, the UK and Singapore (Inc42, July 2026).
• Large enterprises anchor demand: CARPL states it is used by four of the world’s five largest private radiology groups, with reported clients including RadNet (US), I-MED Radiology Network (Australia), Fleury (Brazil), Medica (UK) and Massachusetts General Hospital (US).
• Government and public-health channels: deployments span government initiatives in Brazil, India, Singapore, Spain and the UAE, plus the Clinton Health Access Initiative in India, an unusually public-sector-heavy mix for a young AI firm.
• India as the build base: the company keeps a significant technology-development presence in India while its revenue base skews international, alongside offices reported in the San Francisco Bay Area and Singapore.

The surprise, then, is that a New Delhi company’s marketplace is anchored by some of the largest radiology enterprises outside India, and by public health systems, rather than by the domestic private hospital market.

The risks

• Revenue has not caught up to reach. FY24 operating revenue was ₹10.9 crore (Tracxn) with widening losses (Tofler), while the platform claims 300-plus apps and marquee global clients. If usage and paid deployments do not convert that reach into recurring revenue, the gap between narrative and filings becomes a funding-dependency problem.
• Platform disintermediation. As a vendor-neutral middle layer, CARPL depends on both large AI vendors and large hospital groups. Either side could, in principle, build direct integrations or their own marketplace, squeezing the layer in between, a classic risk for any two-sided platform.
• Regulatory and trust exposure. CARPL operates in clinical decision support across many jurisdictions, each with its own medical-device and data rules. A single high-profile failure of a hosted algorithm, or a data-security incident behind a hospital firewall, would hit the trust the whole model rests on.
• Contested “largest” claim. The “world’s largest radiology AI marketplace” line is company-stated and repeated through CARPL’s own press releases; it is not independently audited, and the copy here treats it as a claim rather than a verified fact.

The takeaway

The transferable lesson from CARPL is about where to stand in a gold rush. When hundreds of companies were racing to build radiology algorithms and over 700 had cleared the FDA, the durable position was not to build the 701st but to own the layer that lets a hospital use any of them safely. CARPL took an internal tool built to solve its own family business’s problem, recognised that the same problem existed everywhere, and sold the plumbing instead of the product. The unfinished part of the story is just as instructive: reach measured in apps and logos is not the same as revenue, and a ₹96 crore Series A backed by the World Bank’s IFC buys time to close that gap, not proof that it is closed. For anyone building infrastructure, the discipline is to keep asking whether supply-side scale is actually turning into paid, repeated use.

Frequently asked questions

What does CARPL.ai do?

CARPL.ai runs a vendor-neutral platform that lets hospitals and radiology groups discover, test, deploy and monitor medical-imaging AI tools through one interface. As of July 2026 it hosted more than 300 radiology AI applications from over 100 partners, and it builds none of those algorithms itself.

Who founded CARPL.ai and when?

It was founded by Dr Vidur Mahajan, a physician with a Wharton MBA who previously ran the family radiology chain Mahajan Imaging. The legal entity dates to 2018 and the platform launched publicly in 2021. The leadership team includes Dr Vasanth Venugopal (CMO), Rohit Takhar (CTO) and Dhruv Sahai (COO).

How much funding has CARPL.ai raised?

Disclosed funding stood at $16 million (about ₹153.6 crore) as of July 2026: a $6 million seed round in February 2024 led by Stellaris Venture Partners, and a $10 million (about ₹96 crore) Series A in July 2026 led by the International Finance Corporation, with Stellaris participating. The valuation was not disclosed.

Is CARPL.ai profitable?

No. It reported operating revenue of ₹10.9 crore for the year ended 31 March 2024 (Tracxn), and Tofler’s data shows losses widened that year, with net profit and EBITDA both falling sharply. It remains a private, loss-making company investing ahead of revenue.

Who are CARPL.ai’s customers?

The company says it is used by four of the world’s five largest private radiology groups, with reported clients including RadNet, I-MED, Fleury, Medica and Massachusetts General Hospital, plus government and public-health deployments in Brazil, India, Singapore, Spain and the UAE.

Sources

Figures are as of September 2026. Currency converted at $1 ≈ ₹96.0 as of 18 September 2026 (Trading Economics).

• TechCrunch, “CARPL guides healthcare providers through the growing market of radiology AI apps,” February 2024
• HIT Consultant, “Radiology AI Marketplace CARPL Secures $6M,” February 2024
• Inc42, “CARPL.ai Bags $10 Mn To Scale Its AI Medical Imaging Marketplace,” July 2026
• YourStory, “Tech30 startup CARPL.ai raises $10M in Series A led by IFC,” July 2026
• Medical Dialogues, “CARPL.ai Raises USD 10 Million in Series A Funding Led by IFC,” July 2026
• BW Disrupt, “Carpl.ai Raises $10 Mn In Series A Round Led By IFC,” July 2026
• Radiology Business, “Radiology artificial intelligence marketplace CARPL.ai raises $10M,” July 2026
• Crunchbase, CARPL.ai company profile (total funding), September 2026
• Tracxn, CARPL.AI Private Limited company profile and financials, September 2026
• Tofler / MCA filing summary, CARPL.AI Private Limited (CIN U73100DL2018PTC333492), FY24, September 2026
• CARPL.ai, official news release on the IFC-led round, July 2026

Found an error? Write to us and we’ll correct it in the open, dated, on the piece.