Bhat's Notes

Excerpts from the Panel Talk: “How to Build a Successful Healthcare Startup” at AIG Hospitals, AI in Healthcare Conference | January 7, 2025

Panel:

1) Sarang Deo of ISB

2) PJ Narayanan, IIIT-G

3) Kshitij Jadhav, IITB

4) Ramesh, LVPEI

Moderator: Rakesh Kalapala, AIG

Rakesh @ AIG: How to Do Research in Healthcare? What’s the Problem to Look For?

PJN: The difficulty of the problem excites us in academia. We are often enamored by the elegance of the solution rather than its real-world impact. Healthcare presents countless problems that startups can solve—problems that may seem uninteresting from an academic lens but deliver significant impact. Addressing these domain-specific challenges is particularly difficult in elite institutions like IIIT, where the focus leans heavily towards new technology.

Kshitij: Talk to the end user. Talk to the operations team. In academia, we often focus on open-source, state-of-the-art solutions. But at Max Healthcare, we needed something as practical as accurate hip measurement—and there was no gold standard available. We used the YOLO algorithm to address it. The key takeaway: the solution doesn’t always need to be fancy. Unless we engage with end users and understand their needs, we can’t create meaningful impact.

Dr. Ramesh: My take on the patients and providers when we take up a solution or device of a startup: The Technology Centre at LVPEI has been there for 10 years, and we always have engineers and doctors together. Engineers are in the hospital, so they work closely with the doctors to solve a problem. We do innovation in India to make the patient’s life and the doctor’s life better. Can we bring an international technology to India? LV Prasad is a pyramidal system. AR systems in villages won’t work due to data bandwidth limitations. We also have an offline solution for this. We started this homecare initiative. We create the ecosystem according to the patient type. We conduct surveys before we build or scale large.

Sarang: The funding options I suggest — I think people in the room who want to start up already know about VCs. The point is about how to convince the investors.

Q) Barriers you usually confront?

PJN: The major challenge is access to data and expertise. Individual data also plays a crucial role in the context of big data. There is often a perception in the medical system that data is proprietary, but in academia, we require data access along with gold standards and ground truth for meaningful research.

Understanding the healthcare domain is far more complex compared to other sectors. While federated learning is still in its nascent stages, I see the primary barrier as gaining deep access to both expertise and high-quality data.

Q) How do you ensure startups integrate into the system and convince both VCs and end users?

Kshitij: Convince the end user, and VCs will follow. Don’t try to disrupt the existing workflows of doctors—they are already overworked and overburdened. Instead, leverage existing deployed systems to integrate your solution seamlessly.

Q) How to recruit talent and retain?

Sarang: You can’t build a healthcare tech startup with just engineers. You need practicing physicians and other experts in the mix. But when will a doctor-founder find time for a startup? Good ones will find the time.

Startups often lack a clear plan for revenue generation. Founders must understand the Indian healthcare system—what constitutes out-of-pocket expenses, payer models, and other financial structures.

They need to identify what drives revenue, how to retain talent, and ensure talented individuals remain actively engaged. Show talented people that the solutions are delivering tangible results and have support systems in place to nurture and retain them.

Hospitals should welcome engineers and management professionals for long-term residency programs to build sustainable collaborations.

Every startup needs someone who can identify the top three problems to solve for an organization. Ask: If I solve the top problem, what will fundamentally change?

Q) How to inculcate the innovation culture?

Ramesh: At LV Prasad, we allocate 25% of our time to innovation, administration, or community health initiatives. There must be a strong collaboration between engineers and doctors, even at the college level.

Colleges often have numerous granted patents with very few being licensed. There are no takers for these innovations. This gap needs to be bridged to ensure meaningful outcomes.

Kshitij: Understand each other’s language.

Q) How to grow and scale?

Sarang:

Innovations often fail to scale either because they are not truly needed or because the path to scaling is unclear. The challenge with scaling up usually lies in achieving marginal improvements using AI that could potentially be achieved without it.

For example, if I have an AI solution to optimize the length of stay for inpatients, the critical questions are:

• What unique value does AI bring to this problem?

• Is this solution truly valuable for AIG Hospitals?

• How much money does it save the hospital?

The reality is, no one has a ready answer to these questions. The real test for a startup is: Can it convincingly demonstrate these value savings to a VC?

PJN)

The key lies in dialogues among communities. Michael Jordan from Berkeley shared an impactful story:

In 1994, his wife became pregnant, and an ultrasound indicated a possibility of Down syndrome. Later, he discovered that the Siemens machine being used was originally built in the 1990s, with only the hardware updated over time. He concluded that the concerning findings were due to sensor noise rather than an actual issue. That same day, similar machines diagnosed a million pregnancies worldwide, and tragically, many babies who underwent follow-up invasive tests did not survive—a loss that was avoidable.

The lesson here is clear: it’s not just about AI—it’s about ensuring the entire system functions reliably and accurately. AI may not always be the ultimate solution, but thoughtful design, rigorous validation, and continuous improvement of technology are critical.

Sarang: Think deeply about what problem needs to be solved. Don’t just bring a solution and try to fit it into this domain. Ask: What is the top problem? And how much improvement in outcomes can we achieve by solving it?

Hospitals should be encouraged to identify and present grand challenges that need solutions.

Kshitij: No one should be captivated by the elegance of an algorithm. You might not even need Gen AI to solve the problem. The key is to identify the right problem and determine whether it can be effectively addressed using existing solutions.

Dr. Ramesh: Collaborate with engineers. Doctors don’t need to become engineers, but they shouldn’t hesitate to ask for help when needed.

Out of every ten doctors, perhaps two could innovate significantly over the next decade. Doctors are naturally skilled at identifying patterns in clinical data.

The real question is: Can pattern recognition technologies reduce a doctor’s effort in identifying these patterns, allowing them to focus on decision-making and patient care?