AI that screens for autism with a webcam: Kamath bets on early detection in rural India

By Zendoric · October 5, 2026.

Indian entrepreneur Nikhil Kamath, co-founder of the stock trading platform Zerodha, has co-led a 4 crore rupee seed round —some $450,000, according to Indian Startup News— in Aignosis, a startup from Jaipur (India) that uses artificial intelligence to screen for autism and other developmental disorders in children between 16 months and 9 years old. The round, co-led with early-stage venture capital firm Antler, also brings in Varun Dua, founder of the digital insurer Acko, and IIMA Ventures, the incubator of the Indian Institute of Management Ahmedabad.

The system needs no special hardware: it runs on a laptop with a webcam. The child watches a nine-segment video while computer vision software —the technology that lets a computer interpret images— records behavioral cues: where the child looks, how they react when called by name, how they respond to facial expressions. The result is a risk report, not a diagnosis: it indicates whether the child should be referred for a full clinical evaluation. Aignosis says it has obtained Class B certification —the low-to-moderate risk category for medical devices— from India's health regulator, the Central Drugs Standard Control Organisation (CDSCO). In 2025, Google selected it for its AI startup accelerator program, out of more than 1,600 applicants from which it chose only 20.

The company, founded by Raksheet Jain and Divyansh Mangal, does not sell the product as a standalone device: it operates under a “clinic within the clinic” model, in which it is the pediatricians themselves who use the tool in their practices. It is already in more than 20 public primary health centers in the state of Rajasthan, has an agreement with the pharmaceutical company Mankind Pharma —whose sales reps help organize screening drives alongside pediatricians— and is now starting with the Apollo SHINE Foundation in some 125 schools. Aignosis says it has carried out more than 8,000 screenings to date and aims to reach 100,000 before the end of India's 2026-27 fiscal year, with expansion planned to Punjab, the Delhi region, Haryana, Tamil Nadu, Jharkhand, Karnataka and Maharashtra.

For Kamath, the appeal is not just the technology but the entry point into the health system. “Most healthcare innovation in India is built for the top of the pyramid and never leaves there,” the investor said. “Aignosis goes in the opposite direction: it starts in public health centers and small clinics, not in private hospitals in the metros.” It is a small round by Kamath's standards; his fortune comes from the growth of Zerodha —India's largest stock trading platform by active clients, built without outside venture capital— and he has become one of the country's most visible individual investors, with stakes in health, consumer and financial services.

What is interesting about Aignosis is not the amount —$450,000 is a modest figure even for a seed round— but the architecture of responsibility it has chosen. By limiting itself to generating a risk score and leaving the diagnosis in the pediatrician's hands, the product sidesteps the problem that holds back much of medical AI: who answers, legally and clinically, when the system gets it wrong. A model that “diagnoses” takes on a responsibility that no regulator yet grants easily; a model that “screens and refers” leaves that responsibility where it has always been, with the doctor, and that is why it achieves regulatory approvals and public health deployments at a speed that diagnostic AI rarely manages.

It is also worth noting what the material does not say: neither Aignosis nor the coverage of the round publishes sensitivity or specificity figures for the algorithm —that is, what share of children with autism it actually detects and how many false positives it generates—. Without that data, published and independently audited, 8,000 screenings and a seed round are a sign of commercial traction, not proof of clinical efficacy. Before the tool reaches the hundreds of thousands of families it has set as its target, that validation will have to be made public.

Even with that caveat, the case points to something bigger than this startup. Early detection of autism substantially improves the results of intervention, and in India —as in much of the developing world, and in the rural areas of rich countries— specialists able to assess it are scarce outside the big cities. If a webcam and a computer vision model manage to put a reasonable screening in the hands of any pediatrician at a rural health center, they do not replace the specialist: they multiply their reach. That is, on a small scale, the underlying promise of AI applied to health that we will keep watching at Zendoric: not to replace medical judgment, but to extend it to those who today have no access to it. The work that remains —regulatory, clinical, independent validation— is exactly the distance between a pilot in 20 clinics in Rajasthan and a national screening network that can truly be trusted.