Upvote for Living Colour. And I had lost track of them too, after their initial breakout success. "Cult of Personality"…
Chakri Toleti is founder and CEO of Care.ai of Orlando, FL.
Tell me about yourself and the company.
This is my fourth gig in the healthcare space. My brother Raj and I have done business together, and this is my own project. I started the business three and a half years ago to bring ambient intelligence to healthcare.
I don’t have a pure technology background or healthcare background. I worked for Disney Ideas, then went to film school. That has nothing to do with any of this stuff, but I always was intrigued with other industries and how they adopt technology to bring process automation and efficiencies to deliver consistent, better solutions. That is my background and my passion.
Looking at healthcare, many processes can be efficiently automated to impact the care delivery process itself. I looked at ambient intelligence and felt that there is a significant gap in healthcare. I saw the transformation that was happening even in your home, like a smart home, the ability to get control of what’s going on in real time. That was the genesis of Care ai.
How will your business change as new types of health-related sensors are developed?
The technology has evolved dramatically. We can deploy high compute engines like GPUs in a smaller form factor with less power consumption. We have several provisional patents in terms of how to scale and do edge computing in a much more efficient way in the healthcare setting. We can roll out across tens of thousands of rooms without bringing the network down. We are really good at being able to get the appropriate data, clean data, to run these AI models on the edge.
If you want to draw parallels, look at Nuance’s 10- or 15-year-old technology with Dragon. When you have enterprises like Google, Microsoft, and Amazon spending billions of dollars on NLP-based workflows, that has become commoditized dramatically. Amazing large language models are being deployed in enterprise settings to be able to deliver the same kind of results, and much better results, for a fraction of the cost. That’s the transformation that is happening.
What we’ve built is bringing these operational clinical workflows together, building a scalable command center, and shifting the paradigm of what clinical data capture or operational data capture will look like in healthcare.
A lot of the old-school monitoring in the ICU went beyond sensor-based instrument alarms and instead involved an experienced clinician asking questions or observing the patient. Can value be added by analyzing audio and visual information?
That’s exactly what we do. Imagine a Tesla car sitting in a room. That’s what we’ve built — inferencing, audio-visual, three-dimensional volumetric data to give you a lot more information of what’s going on, how many people are in the room, how long did they stay there, did the patient eat food, how long have they been sleeping in the same position. All the environmental data, coupled with the data capture of every action that’s happening, is the fundamental difference that we are enabling to truly build a smart patient room.
I wake up every day from the dream that I’m going to kill the EMR. EMRs are the most antiquated way of data capture. They are required, but were built for a specific purpose 10, 15, or 20 years ago with an archaic way of data capture. It would be unthinkable if workers in an Amazon warehouse had to stop and input information about everything that they are doing. Yet we take the most talented and expensive resources in healthcare and make them do data entry in a crappy interface with all these clicks, forms, and flows in a complex form of data capture. All it is doing is generating a bill.
Obviously the clinical data is important, but we all know that every unit in every health system has skewed dark data. If you look at the respiratory rate, it is magically the same, 14 or something, in every unit. It’s like muscle memory. It gets worsen as you go through the ecosystem. Post-acute reimbursement is completely based on data capture. They have something called ADLs, activities of daily living. They have to capture all of that, and it’s a manual process.
Some hospitals have created command centers and are interested in remote patient monitoring. What will the hospital of the future look like given the opportunity to separate the services from the hardware capabilities of the room or having people enter the room regularly?
An accelerator for us is that the staffing shortage and the staffing crisis is elevating the need for solutions like these that can give the bedside care teams the scale that they need. Also, they have to think outside the box. The EMR cannot be the universe of every way of capturing information. Every health system recognizes that, and that’s why we are getting traction.
Also, the technology has become democratized, in that the cost to deploy these solutions is fractional. If you go to most of these organizations, they are still moving computers-on-wheels from older companies from one room to another, paying $10,000 or $30,000 per cart. For a fraction of that cost, we can wire up a true smart patient room that gives you real-time visibility into operational and clinical workflows with the ability to analyze audio, video, three-dimensional volumetric data visualization and capture of that information with super high accuracy.
How will AI change the way we think about healthcare software and how technology is developed or deployed?
It will be a once in a generation change in terms of how you look at delivering care. There are two sides to it. One is innovation, drug discovery and all the other aspects of AI. But when it comes to the four walls of operations of a hospital or post-acute facility itself, real-time AI will fundamentally change how we monitor and how we deliver care in an efficient way and at higher standards of quality. If you look at generative AI and all the innovation that is happening at an accelerated rate, healthcare will have a huge impact on that.
When we talk about AI in a healthcare setting, people talk about taking a few algorithms and applying them to the dataset that we have. That is good, and you need it. But a lot of the data is dark data. It’s skewed. How did we capture that information? Is it accurate? You have to go back and look at how you bring true, clean data into the system.
Imagine a self-driving car. They send out these cars, capture real-time information about the roads, then teach the neural nets to look for the most efficient way of driving. More and more you will see those kind of implementations and adoption of AI into healthcare in a different way. It could be a radiology or a CT scan that’s happening in real time. The ability for it to recalibrate itself using AI to get more accurate scans will also be a part of the entire ecosystem. Rather than just, hey, I’ve scanned, so let’s apply AI to identify abnormalities. There are different aspects of AI that have not fully been leveraged in healthcare settings.
How should a mid-sized healthcare technology company look at incorporating large language models that are changing so quickly?
We should be looking at a problem and then seeing if applying AI to that problem will solve it. Does it even require AI? Once you have identified a problem like nursing shortages — we have a virtual nursing infrastructure — but then how do you look at AI being more integrated into the platform? Understanding the workflows within healthcare and using the frameworks with the right set of data to impact that workflow. That work will be a key way for these organizations to succeed.
Cerner or Epic were designed before a lot of these innovations happened. For example, for controlled substances, two people have to sign off in the room, logged into the EHR on the same computer. That was designed like 10 years ago. There’s no way for one person to be virtually beaming in and one person in the room. EMRs don’t have the ability to do it. They would have to re-architect everything in the new way of doing things. That would be a big lift for them.
Newer companies have an advantage to look at a clean slate and say, what’s the most effective way in today’s technology landscape to implement the most effective solution for that problem? If they truly understand what real-time AI can do, then the sky’s the limit to transform healthcare.
You started the company right before the pandemic began. What is different now about starting, running, and selling a digital health company?
I would strike out the last one. If someone is building something with the objective of selling it, then that’s the wrong way of going about it. You have to solve a problem, and whatever the outcome is, it will be good, whether you sell the company or stay with it.
The landscape has dramatically changed. For us, we had an advantage in that we started the business when the pandemic hit, which propelled and accelerated our growth. I don’t think I could repeat the same kind of growth again in my career. We were at that inflection point.
Also, health systems have changed their thought process. The pandemic exposed the weaknesses that are inherent in the care delivery system and processes. That is in the forefront of the leaders in these health systems for them to solve. They are much more open to new, innovative companies, so it’s a great time to bring innovative technologies to these institutions that are more open to newer ideas and newer companies to innovate for them. They know that the status quo has a lot of weaknesses that are built into their systems today. It’s a great time if you have the right solution to help them be more efficient and deliver the same or higher standards of care.
What will be key to the company’s strategy in the next three or four years?
It will be extremely important to understand the impact of AI and how it will change the client’s businesses. If companies don’t look at new ways to solve problems, be nimble about it, and adapt aggressively, it will be tough in a dynamic environment. The technology landscape is changing at a much faster pace than we’ve ever seen in our careers. They have to be at the same speed as what the technology is changing. ChatGPT 3.5 versus ChatGPT 4 or Bard are coming up at lightning speed, and startups and new companies that are trying to go to market need to have the same agility.