Dr. Nakul Shekhawat and Jordan Shuff Are Using AI to Restore Sight

Seeing shouldn't be a luxury. According to Dr. Nakul Shekhawat and Jordan Shuff, smartphones might hold the key to making this belief a reality. They co-founded Visilant to use AI-powered mobile tools to diagnose eye disease in low-resource communities across India. They’re proving that AI can expand access to sight-saving care in places where doctors and clinics are often out of reach.

Nakul is an ophthalmologist who began his career delivering humanitarian eye care in rural India. Jordan is a biomedical engineer turned AI technologist who has spent seven years leading global health projects. Together, they’re unlocking a new AI-powered model for accessible vision care.

In this conversation, they share how real-world data shaped their AI system, why human connection still matters in digital diagnosis, and what it takes to build health AI tools with equity at the center.

How did your journey inspire you to explore AI for humanity?

Nakul: Nearly two decades ago, I volunteered for a vision screening program in rural India. It was close to where my grandparents were born and raised. Our team of eye doctors drove hours to a small desert village. There, hundreds of patients were waiting to be screened for blinding eye diseases, hoping to be selected for sight-restoring surgery. We could only visit that village a few times a year. Desperate patients fought each other to reach the front of the line, terrified they'd miss their chance to regain their vision.

Our bus only had seats for 16 people, but we managed to squeeze 61 blind patients onto our bus and take them back to the city to receive sight-restoring cataract surgery. These patients had to be led in by hand because they were functionally blind. After surgery, they could walk, see, and regain the independence that had eluded them for years. That was the day I decided to become an ophthalmologist.

There have been no recent major innovations guiding how we screen patients for commonly blinding eye conditions in low-resource settings. Doctors still have to travel long distances with bulky equipment, and far too many patients slip through the cracks. In fact, 1B people around the world live with untreated vision loss.

As an ophthalmologist, I can examine only a limited number of patients in a day. But AI can examine thousands. If we can use AI to remotely diagnose patients earlier and guide them to care, then doctors like me can focus our human expertise on actually treating patients to restore vision.

Jordan: In 2020, as part of my biomedical engineering master’s degree at Johns Hopkins University, I was tasked with finding the biggest unmet need in ophthalmology and building a solution. Sounds easy, right? I landed with a one-way ticket at Aravind Eye Hospital in Tamil Nadu, India, just as the pandemic shut down in-person eye screenings. Villagers began texting blurry photos of their eyes, and even their neighbors’ eyes, to doctors, hoping to get their blinding conditions diagnosed. This community-led momentum inspired me to build a system that made it easy for villagers to capture clinical-grade eye images on their smartphones and securely send them to the hospital from anywhere.

As more images came in, we needed more doctors to review the photos than we needed to treat patients. So we tested a basic deep learning model to classify images by diagnosis. To our surprise, it worked remarkably well. Despite my extensive technical background in medical imaging analysis, I hadn’t planned to work on AI. But AI turned out to be the right solution at the right time for our biggest unmet need. Since then, it has helped doctors remotely diagnose patients at high volume, maintain accuracy, and ensure vulnerable patients don’t get overlooked.

How did you train your AI to make accurate diagnoses from just smartphone images, even in the most remote places?

Nakul: We’re building AI that can diagnose and triage the leading causes of blindness. It uses smartphone eye images and basic clinical data like vision tests and symptoms to guide care decisions. Our models run directly on a smartphone, even without internet. Through AI, we empower non-specialists like community health workers to diagnose eye diseases, determine if referral for in-person eye exams is needed, and decide how urgently the patient needs care.

We designed the AI to support practical clinical decision-making, not just identify diseases. It’s not helpful to simply tell a patient they have cataracts. You have to determine if the condition is advanced enough for surgery, and whether the patient faces clinical risk factors or barriers to care that might prevent successful surgery. That level of triage requires a deep understanding of clinical endpoints and workflows so the AI can guide patients to the right care. We worked closely with doctors and hospital leadership to map out referral and treatment pathways for each disease presentation, ensuring the model’s outputs aligned with how actual clinical decisions are made on the ground.

Jordan: To ensure our algorithms performed well in real-world conditions, we focused on capturing real-world images, rather than perfect clinical images obtained in controlled settings. All of our training data comes from budget smartphones used by health workers in the field. We built a coalition of hospitals across India that believe in this technology. With their partnership, we’ve collected a set of geographically diverse, high-quality images. Most of the work isn't glamorous. It involves carefully reviewing and labeling every image that comes in and incorporating it into what is now one of the largest anterior eye training datasets in the world.

Because our models are built to run offline directly on smartphones, internet access is not a barrier to care. This meant investing in model optimization and deeply understanding clinical endpoints so that we could reduce model size without sacrificing the most important diagnostic and triage decisions.

What lessons have you learned about balancing AI automation with the critical human touch in healthcare?

Nakul: As an ophthalmologist, clinical safety is my top priority, so we built clear guardrails. We loop in ophthalmologists any time the model is uncertain, the patient needs surgery or medication, or the patient reports symptoms that the model cannot address. For example, if the AI determines a patient’s eyes appear healthy, but the patient says they’re still worried about their eye health, they deserve to speak with a doctor. No one should walk away feeling dismissed.

Jordan: Primary care doctors and community health workers use our AI models with patients in the field. We’ve found that having a human in the loop builds trust for both the provider and the patient, and leads to better eye screening outcomes. These health workers not only operate the AI tool, but they also ensure patients understand their results, explain the importance of follow-up care, and reinforce the need for timely referrals. Since many of these health workers are trusted community members, their involvement helps build rapport and ensures that patients feel supported throughout the process.

What core values drive your unique vision for impact in an AI-driven future?

Nakul: Too often, innovation benefits high-income settings first, trickling down slowly to underserved communities. We're inverting that model. By building imaging tools and AI algorithms designed to perform where the clinical need is the greatest, we’re demonstrating that technological excellence and equitable access are not mutually exclusive. They can and should go hand in hand from day one.

Jordan: I believe AI works best when it increases human agency rather than replacing human connection. In 2020, we saw the beginnings of this potential when community members began texting eye images to doctors at Aravind. Today, our AI empowers community health workers to deliver care locally and educates patients about their eye conditions, which sparks conversation and reduces fear of the unknown. When patients understand what’s happening with their eye health, they feel empowered to make decisions about their care with confidence and dignity.

Which visionary leaders, philosophies, or movements give you hope for a more human-centered AI future?

Both: Aravind Eye Hospital!

Nakul: Aravind is a pioneer in making advanced technologies accessible to everyone regardless of location, wealth, or status. Their globally recognized approach to scaling impact through affordability and access is a model for how AI can be used to bring critical healthcare solutions to the most underserved populations, and it’s exactly the kind of vision that excites me about the future.

Jordan: Hard agree! In a system with too few doctors, many rural patients, and limited resources, Aravind has managed to provide world-class treatment to millions of patients free of charge. They turned the system’s constraints into fuel for innovation. By training thousands of rural women to perform care tasks, they created a highly efficient model that delivers exceptional care at low cost without relying solely on doctors. What gives me hope is that their model is not only technically brilliant but also rooted in the belief that healthcare must serve the poorest patients to truly work. To me, a human-centered AI future is one where a deep commitment to inclusion shapes both how we build and where we choose to innovate.

What is your 7-word autobiography?

Nakul: Ophthalmologist; global health innovator; pragmatic AI optimist.

Jordan: Curious builder. Vivacious storyteller. Persevering towards vision.