Sunday, September 25, 2022
HomeRoboticsVarun Ganapathi, CTO & Co-Founder of AKASA - Interview Series

Varun Ganapathi, CTO & Co-Founder of AKASA – Interview Series

[ad_1]

Varun Ganapathi is the CTO and Co-Founder of AKASA, a developer of AI for healthcare applications. AKASA helps healthcare organizations improve operations, including revenue cycle, to drive revenue, create efficiencies, and enhance the patient experience. Varun has successfully started two AI companies prior to AKASA, one was acquired by Google and the other by Udacity.

You’ve had a distinguished career in machine learning, could you discuss some of your early days at Stanford when you worked on making helicopters autonomous?

When I was studying physics as an undergrad at Stanford, I was also very interested in computer science and machine learning (ML). To me, AI and ML combined everything in one – it’s really an automated way of doing physics on any digitizable phenomena.

For this one particular project, we had this helicopter that looked like a large drone a bit smaller than a twin mattress – at a time when drones were not prevalent. People were flying it and making it do tricks, such as hovering upside down. While this is very difficult to do, we wanted to build an ML algorithm that could learn from humans how to fly this helicopter autonomously.

We created a physics simulator that was based on the actual helicopter and an ML algorithm that learned how to predict its movements. We then applied reinforcement learning within the simulator to develop a controller, took the software, and uploaded it into the actual helicopter. After we turned the helicopter on, it worked on the first try! The helicopter was able to immediately hover upside down by itself, which was pretty impressive. The team continued to work on automating other types of tricks using ML.

You also worked at Google Books, could you discuss the algorithm that you worked on and how your company was eventually acquired by Google?

I actually did an internship at Google while taking classes at Stanford in 2004 – this was right after the helicopter project. During that time, I was implementing ML for the Google Books project where we were scanning all of the world’s books.

Google was paying all these people to label information about the books, such as pages, tables of content, copyright, etc. – a very time-consuming task. I wanted to see if we can use ML to do this and it worked really well. It actually performed better and was more accurate than when humans did it because most of the errors were due to human error with manual labeling.

This got me really excited about ML because it showed that you can go from human performance to superhuman performance – doing mundane tasks with fewer errors and more consistently while still handling edge cases.

From there, I decided to do a Ph.D. at Stanford, focusing on ML and more theoretical papers at first. For my thesis, I developed an algorithm to perform real-time motion capture where a computer can track the motion of all human joints in real time from a depth camera. This was the basis for my first company, Numovis, which focused on motion tracking and computer vision for user interaction. It was acquired by Google.

My entire journey from the helicopter project to Google Books to self-driving cars and now healthcare operations really showed me how powerful and general machine learning algorithms are.

Could you share the genesis story behind AKASA?

We’ve built AKASA to fix a massive, deeply embedded problem in healthcare operations. These operations are both expensive and error-prone which can lead to unnecessary panic-inducing financial experiences for patients. There was a lack of new technology on the administrative side and nothing being purpose-built. It became clear to us that you could use technology like AI and ML to solve these operational challenges in an innovative way. When we spoke to a multitude of health systems and healthcare leaders, they validated our thinking which ultimately led to the foundation of AKASA in 2019.

With that, AKASA’s purpose has been clear from the beginning – to enable human health and build the future of healthcare with AI. The way we decided to take on this challenge is by combining human intelligence with leading-edge AI and ML so health systems can reduce operating costs and allocate resources where they matter most.

Our system-agnostic, flexible platform is currently serving a customer base representing more than 475 hospitals and health systems and more than 8,000 outpatient facilities, across all 50 states. Our technology helps these organizations whether they’re using electronic health record (EHR) providers like Epic, Cerner, other EHRs, or bolt-on systems, and everything in between. And we’ve done it with strong results.

Our customer base represents more than $110 billion in aggregate net patient revenue, which equates to more than 10% of all U.S. health system spending annually according to the Centers for Medicaid and Medicare Services. And AKASA’s models and algorithms have been trained on nearly 290 million claims and remittances.

The invisible plumbing of healthcare is extremely complex, but it has an immense impact on human health, and we’re automating it bit by bit.

What are some of the tasks that AKASA is looking at automating in healthcare?

Our unique expert-in-the-loop approach, Unified Automation™, combines ML with human judgment and subject matter expertise to provide robust and resilient automation for healthcare operations. AKASA can quickly and efficiently automate and streamline end-to-end tasks within the healthcare finance function, including bill processing and payments. Specific tasks AKASA automates include checking patient eligibility, documenting and verifying insurance information, estimating patient cost, editing, rebilling, and appealing claims, and predicting and managing denials.

This type of automation not only reduces human error and delays for patients, helping prevent surprise medical bills, but also frees up healthcare staff by taking the manual, repetitive tasks entirely off their plate – allowing them to focus on more rewarding, challenging, and value-generating tasks directed towards the patient experience.

What are the different types of machine learning algorithms that are used?

AKASA uses the same machine learning approaches that made self-driving cars possible to provide health systems with a single solution for automating healthcare operations. This approach – centered around ML – expands the capabilities of automation to take on more complex work at scale.

We develop state-of-the-art algorithms across computer vision, natural language understanding, and structured data problems. Our platform starts with computer vision-powered RPA and enhances it with modern AI, ML, and an expert-in-the-loop to provide robust automation.

To provide a high-level overview of how it works, our proprietary solution first observes how healthcare staff completes their tasks. Our team then labels that data and uses it to train our algorithms so our technology can understand and learn how healthcare staff and their systems work. From there, our platform performs those workflows autonomously. Finally, we use experts-in-the-loop who can jump in whenever the system flags outliers or exceptions. The AI continuously learns from those experiences, allowing it to take on more complex tasks over time.

Could you discuss the importance of human-in-the-loop approaches and why this is set to displace RPA?

The hard truth is that RPA is a decades-old technology that is brittle with real limits to its capabilities. It will always have some value in automating work that is simple, discrete, and linear. However, the reason automation efforts often fall short of their aspirations is because life is complex and always changing.

The basic approach to RPA is building a robot (bot) for each problem or path that you want to solve. A human (consultant or engineer) builds a robot to solve a specific problem. This robotic solution takes the place of a sequence of steps. It looks at a screen, takes action, and repeats it.

The problem that often occurs is that a change in the world, such as a modification to a piece of software or UI, can cause bots to break. As we know, technology is ever-evolving, creating dynamic environments. This means that RPA robots often fail.

Another problem with these bots is that you need to create one for every situation you want to solve. Doing this, you end up with many robots, all completing very small actions that don’t require much skill.

It’s like a game of whack-a-mole. Every day you face the likelihood that one of them will break because a piece of software is going to change or something unusual will happen – a dialogue box will pop up or a new sort of input will occur. The result is costly maintenance to keep these bots running. According to research from Forrester, for every $1 spent on RPA, an additional $3.41 is spent on consulting resources.

In other words, the actual software for RPA is not the majority of the cost. The more considerable cost investment is all of the work that you have to do to keep RPA running all the time. Many organizations don’t account for that ongoing cost.

As so much of life is complex and constantly evolving, a lot of work falls outside of the capabilities of RPA, which is where ML comes in. ML enables us to automate the hard stuff. And we believe the special sauce is humans who improve the algorithms by teaching them.

When the algorithm isn’t sure about what it should do (low confidence), it’s escalated to a human-in-the-loop instead. The humans label those examples and identify cases not handled by the current model. When this is done, and the AI got it right, that’s a well-functioning task.

Every task where a human catches a problem is a case where the machine isn’t handling it properly. In this case, data is added to our data set, which retrains the ML models to handle this new situation.

Over time, the ML model builds resilience to these new edge cases. This results in a system that is robust and flexible to new outliers or exceptions, and the system gets stronger with time. This means the automation gets better and better and human intervention will decline over time.

Having human experts in the loop is critical to making AI smarter, faster, and better. We need humans to properly train the AI and ensure that it can handle the outliers that are an inevitable part of any industry – and especially in a dynamic field like healthcare.

How does AKASA’s human-in-the-loop solution Unified Automation™ work, and what are some of the primary use cases for this platform?

Unified Automation is a platform purpose-built for healthcare. Using AI, ML, and our team of medical billing experts, it creates a seamlessly integrated, customized solution that helps you see value faster, with virtually no maintenance or exception queues.

It has been designed with exceptions and outliers in mind. If it encounters something new, the platform flags the issue to AKASA’s team of experts who resolve it while the system learns from the actions they take. It’s that human element that differentiates us from other solutions in the market and allows the platform to continuously learn and improve.

Unified Automation also adapts to the healthcare industry’s dynamic nature. It’s a seamlessly integrated, customized solution that helps reduce operating costs, elevates staff to tackle more rewarding work which requires a human touch, and improves revenue capture for health systems while also improving the patient financial experience.

Here is how Unified Automation works:

Proprietary software observes: Our Worklogger™ tool remotely observes how healthcare staff completes their tasks. Then our team labels that data and feeds it into our automation to provide a comprehensive view of current workflows and processes. This results in higher visibility into staff performance, foundational data on the workflows to power our automation, and an accurate time-per-task analysis.

AI performs: After observing and learning the healthcare staff’s workflows, our AI then performs these tasks autonomously. It continuously learns from problems and edge cases it runs into, taking on more complex tasks over time. Unified Automation sits upstream in the work queue – assigning itself applicable tasks and completing them without disrupting the team. It also automatically optimizes processes so no set-up or intervention is needed from staff.

Human expertise ensures:  The system automatically flags our team of medical billing experts to handle exceptions and outliers, training the AI in real-time as they work. This is the expert-in-the-loop part. With continuous learning built in, the Unified Automation platform gets smarter and more efficient over time and the work always gets done.

Is there anything else that you would like to share about AKASA?

We have a research-first approach which means that our customers have access to leading-edge technology. We are committed to publishing our AI and approaches in peer-reviewed publications to continually set new state-of-the-art standards for AI in healthcare operations and to lead our entire industry forward.

For example, our research has been presented at the International Conference on Machine Learning (ICML), the Natural Language Processing (NLP) Summit, and the Machine Learning for Healthcare Conference (MLHC), among others. We’re taking a very disciplined approach to testing our models and comparing the performance against state-of-the-art AI approaches on the market.

Our predictive denials solution is believed to be the first published deep-learning-based system that can accurately predict medical claim denials by more than 22% compared to existing baselines. Our Read, Attend, Code model for the autonomous coding of medical claims from clinical notes has been recognized as defining a new state-of-the-art for the industry and outperformed current models by 18% – surpassing the productivity of human coders. We believe these back-office innovations are critical to improving the U.S. healthcare system at scale and will continue to drive advancements and build customized solutions for this space.

There’s a lot of hype around AI in healthcare but when it comes down to it, companies can overhype what their technology can actually do. It’s a lot harder to conduct research to validate what the algorithms do – and we pride ourselves for taking this meaningful, yet challenging route to ultimately prove that AKASA’s Unified Automation platform is truly bringing positive and meaningful change to hospitals and health systems.

We’re excited about the future and what’s to come at AKASA as we build the future of healthcare with AI.

Thank you for the great interview, readers who wish to learn more should visit AKASA.

[ad_2]

Source link

RELATED ARTICLES

LEAVE A REPLY

Please enter your comment!
Please enter your name here

Most Popular

Recent Comments