Dell Technologies World 2024 | Kris Kristinsdóttir | Northwestern Medicine

Clips
News
More from Dell Technologies World 2024

Kristjana Ósk Kristinsdóttir

Quantitative Research Engineer

Northwestern Medicine

Generative AI in healthcare: Northwestern aims for global rollout of AI radiology

The integration of generative AI in healthcare is pushing the boundaries of medical innovation.With advanced computing power and strategic collaborations, healthcare providers are beginning to harness these technologies to achieve unprecedented efficiency and improved patient outcomes. One such example is the application of multimodal generative AI models in radiology, which has demonstrated significant improvements in performance and quality of care, according to Kris Kristinsdóttir (pictured), machine learning engineer associate at Northwestern Medicine.Northwest Medicine’s Kris Kristinsdóttir talks to theCUBE about generative AI in healthcare

play_circle_outline Overview of Northwestern Medicine's research and development team

play_circle_outline Revolutionizing Healthcare: Unleashing the Potential of Generative AI in X-ray Imaging for Report Generation

play_circle_outline Mention of future projects and the potential for a "digital twin" in healthcare

play_circle_outline Discussion of the potential impact of AI in healthcare in the future

Info

Kris Kristinsdóttir | Northwestern Medicine

Kristjana Ósk Kristinsdóttir

Quantitative Research Engineer Northwestern Medicine

search

>> Good morning, AI fans and welcome back to fabulous Las Vegas Nevada. We're here on day three of Dell Tech World. It has been an awesome and insightful week. I feel smarter than when I woke up on Monday. My name's Savannah Peterson. You're watching theCUBE. Very delighted to be joined by John Furrier today. Good morning, John.

>> Great to see you. Great event, amazing content. The guests have been phenomenal. Content quality-

>> Yes....

>> has been amazing. Generative AI is here and we've got great use case examples and it's real.

>> Yes, and speaking of generative AI and real use cases, very excited to welcome to the stage Kristen . So, nice to have you here. You're a quantitative researcher at Northwestern Medicine. Welcome, first of all.

>> Thank you guys for having me.

>> How's your week been?

>> It's been great. I've already learned a lot. It's a great conference. It's my first time in Vegas.

>> Oh, my gosh.

>> So, quite overwhelming, but also very fun.

>> Oh, my gosh. Wow. First time in Vegas. John, do you remember your first time in Vegas?

>> Oh, God, no. I can't, actually. I've been here so many times. No, it's like, this is our home. We live here.

>> I was going to say.

>> We live in the Venetian. I could walk blind in my room. The tissues are over there. No, we love coming here. We love the stories. Dell's been great too. I mean, you're working with Dell Innovation Lab, right? HPC Innovation Lab?

>> We have. We've been working with them a lot. They've gotten some great help. They're a great partner in the sense where they help us kind of find the technology that we need. There's so many solutions out there, so they were able to help us kind of puzzle together what would work best for us and set up a great system.

>> Kris, I got to ask you, because one of the things we've been looking for, last year when we were here, we said next year is about workloads in production, real use cases that have real impact. Societal impact, human impact, business productivity, all across the board. This is this year. You guys are doing something really cool with generative AI that is not only cool technically, it's got training and inference, and you've got HPC in there. So, it's very nerdy, which is great. We'd love that. We love the nerd talk, but also you're doing something really compelling that saves lives too. So, I want you to share your story, tell us what you're working on and take us through that.

>> Absolutely, will do. So, I work with the research and development team at Northwestern Medicine. Now, Northwestern Medicine is a hospital system in the Chicagoland area. We have 11 hospitals and over 200 sites of care. And the research and development team is a group of 20 people who have backgrounds in engineering, data science, medicine, and nursing. And we all come together in our mission to try to advance technology within the healthcare system and improve patient care. So, we've worked with a lot of different areas. We've worked in women's health, we've worked with wound care nurses and radiology space. And our projects range from anything from being able to diagnose a disease to optimizing workflows and reducing physician burnout. So, one of our more exciting projects right now is actually a multimodal generative AI model that's able to look at an X-ray image and generate a report. And we have this model live in our hospital system. We have 50 radiologists using it since this summer, and we're seeing some incredible performance improvements. The radiologists are 40% more efficient, and they're also telling us, they're giving us feedback like, "Now we get to spend more time doing what we actually love to do."

>> Helping people.

>> Helping people.

>> Take us through the workflow and what the technology is. You've got data, you're scanning X-ray scans in, you're training it and you're making inference. Take us through some of the details under the hood.

>> Right. We actually have our model running on all x-rays that are taken within our hospital, that radiologists use a system to write the reports called PowerScribe. And in that system you have different templates based on your preference and our model, the ARIES system is just another template. So, the radiologists can choose to use this pre-generated report. They then look through it, make sure everything looks correct before they send it to the doctors and to the patient. So, that's how it looks like. We used to train everything on the cloud before. We drove all innovation on the cloud, and we still do for a lot of our projects. There's multiple use cases that the cloud works very well for. But what we were finding, we're finding ourselves as we grew and as we scaled up, we're finding ourselves more and more limited by our compute resources. Now the cloud is very rigid and also our data lives on-prem, so having to funnel all that data from our on-prem to a GPU cloud instance was becoming more and more of a bottleneck in our training. We got some incredible help from the Dell HPC Innovation Lab. They helped us kind of set up and puzzle together what would help us and what we now have are four PowerEdge XE9680 servers that are all on-prem and HIPAA-compliant. And we have eight Nvidia H100 GPUs per server. And we're seeing some incredible runtime improvements with this new setup. I can give you an example.

>> Please do.

>> Yeah, we'd love it.

>> When we first trained our model, this generative AI radiology model called ARIES, we use a sequence length of 40-96. So, a sequence length is the number of tokens that you can process at the same time. Now, when we trained this with a sequence length of 40-96, one EPOCH took eight days to train. What we've done now is we've scaled up to a sequence length of 200,000. So, by doing that, we're not only able to increase our resolution of our training data, we're also able to do training on much bigger studies. So, what we're doing now is we're expanding to CT.

>> Amazing.

>> Now, CT is 200 x-rays, images, and four different views, so there's so much more data. So, scaling this sequence length up is very helpful for our use cases, and we're able to do it in the same time as we did with the old ARIES model, the x-ray model. So using this new setup, we're able to go from a sequence like the 40-96 to 200,000, which means that we're doing 50 times more backward passes in our training, and we're able to train it in the same amount of time. So, we're in fact seeing a 50-fold increase in our runtime.

>> Oh, my gosh.

>> Which is incredibly helpful. And we're so grateful to the Dell AI team for helping us make this possible and make real impact.

>> Well, I mean Dell in medicine are also very synonymous. Michael Dell has a huge passion for that. He founded their own medical school. It's no surprise that Dell would be a partner wanting to help see this. So, now that you're able to do this better and faster, is Northwestern Medicine going to roll this out to other hospitals and other radiologists around the country, around the world?

>> It's definitely something we're thinking about, absolutely. We want to be able to help as many people as we can. But another vision we have, and another thing we would like to see is that every hospital has their own AI team and their own AI factory. I mean, the healthcare and life sciences industry is responsible for generating more than 30% of all data generated globally. And within our hospitals, we're only using 3% of that data. So, there's so much data that we're not leveraging that we can use to help us solve problems.

>> You know what's interesting, is that you think about the impact of getting that training set bigger and look at the different CT scans increases the aperture of what you can do. At the end of the day, you have customers, end users. They have to come in for an appointment and then a technician has to shoot the film, whatever the technical term is for taking the scan. They got to do the scan and then they got to process it, send that to someone to look at. The person goes home. Someone might not be on call, then they got to come back to do another angle maybe. So, in the workflow, there's a lot of inefficiencies.

>> Absolutely, absolutely.

>> And hassles for the customer, the person who might get overlooked.

>> Yeah.

>> So, I can see, how much do you guys put into that thought and can you share what the impact will be as this rolls out?

>> Absolutely. I can actually give you a good example. On top of this ARIES system, we recently added a feature that is able to prioritize the radiologist's cue to put more severe cases on top of the cue, so they're read quicker. And we just a week, I think, after we deployed this prioritization model, we had a patient come in on a Friday afternoon and ARIES was able to spot that this patient had a pneumothorax, which is a very severe condition, has to be treated immediately. Now, what sometimes happens on Fridays is that the radiology reads get carried on through the weekend. So, the radiologist actually won't look at it until the coming on Monday. But with our algorithm, we were able to prioritize that patient, make sure their scan got read by a radiologist, and that our clinical teams were able to intervene, which might not have happened if we didn't have it.

>> Amazing. And that has real tangible impact on that patient's life.

>> It does.

>> And also probably makes that radiologist feel better that it didn't go over the weekend.

>> Absolutely. Yeah.

>> The workflow of the practitioner, the radiologist, it's in the template.

>> Yeah.

>> So, you guys on the back end, so the back end, what's next? What's next with Dell? Because you get more horsepower, you have a lot of GPUs. Congratulations. That's tough to get those these days. It's always fun to get-

>> Yeah, that's a good point....

>> the GPUs, devours a lot of data. What's next?

>> That's a very good question. We have some very exciting projects in the pipelines. We're working a lot with text now and the medical record and being able to feed the whole medical record into a language model and do all sorts of different stuff. I think our long-term vision is, I mean, the healthcare system is very, very complex. At the core of it, we have the human body, which is by itself very intricate. We have more than 30 trillion cells and even more chemical reactions happening every second. Now at the hospital, we have all these sensors and lab tests and imaging studies that add to that complexity. And then running a whole hospital system and taking care of people is also super complex. You have all the billing and you have managing beds. We thought, what if we can model this? What if we can actually create some sort of digital twin for the system that will enable us to say, "Okay, if this part looks like this and we put it in this environment, what's the outcome going to be? What is the future of this patient? Can we predict future disease states? Can we predict response to treatment and provide a more personalized care?" For example, suggest an optimized diet. So, that's something we're very interested in.

>> You're pointing out a really good thing that we talk about a lot on theCUBE. When people think about multimodal foundation models, think about large language model, text, multiple languages, whatever, images videos. But healthcare is all about multimodal. Scans, DNA's a mode, you know, all kinds of things are now different modes and not text.

>> Right. -

>> Gen AI is a dream scenario for healthcare.

>> It is. It really is. I think healthcare is such a sweet spot for AI.

>> It really is.

>> And it's amazing.

>> You're excited, aren't you?

>> I am. I'm really excited and optimistic.

>> I'm excited now. I want to work there.

>> No, I think it's fantastic. So, when you were working with the team, how did you decide to tackle this problem first?

>> That's a very good question. We have an amazing director to our team, Dr. Mozzi Etemadi, who is a anesthesiologist himself, but also very knowledgeable about AI and generative AI, and he kind of helps us funnel the projects we want to tackle and think about what has the biggest impact. Another thing we are very focused on is getting clinical buy-in from the get-go. We don't go into a project without having a clinician champion to support us because it's very important that we're not creating something that we think will help, but then doesn't. That we actually work with our clinicians to kind of shape the future of AI.

>> That's awesome. Okay. Speaking of, now I'm curious because you're clearly very brilliant and doing some awesome work. What's your background? How did you end up doing this?

>> Oh, that's a very interesting question. I'm Icelandic. I'm born and raised in Iceland. I have a background in biomedical engineering, and after I graduated, I worked as a key account manager for a medical device distributor for a while where I got to spend a lot of time with clinicians discussing technology and integrating technology. I was so surprised by the state of technology within the hospital. After having learned about all the different applications of AI and healthcare and all this shiny technology, clinicians were just frustrated with technology. So, on my mission to try to get healthcare technology up to date, I moved here to the States, did my master's in bioengineering with an emphasis on AI in healthcare at UC Berkeley in California, and then got word of-

>> Go Bears.

>> Go Bears.

>> Absolutely Bears.

>> Yeah. Then made the move to Chicago. Got word of the research and development team. I've always been very interested in being able to work in-house alongside our clinicians to have the most impact.

>> Wow. What a journey.

>> Well, congratulations. You got a great mission.

>> How cool.

>> Love what you do.

>> Thank you so much.

>> We need more healthcare technology.

>> Is she our first Icelandic guest?

>> Yeah. I think.

>> I think she might be our first Icelandic guest.

>> Amazing.

>> You are representing Iceland.

>> Amazing.

>> Is there anyone back home you might want to say hi to in Iceland?

>> Absolutely. Everyone. My parents, my boyfriend. Yeah, everyone's home.

>> All right.

>> Oh, that's fantastic. Wow. Well, I really can't wait to see what you do next. I'm going to ask you one final question. Let's look way into the future. I mean, I love that you were thinking about AI and healthcare years ago, if this is what you were getting your education in, clearly someone with a lot of foresight. 20 years from now, what sort of healthcare challenges, and there's obviously a lot, but what's your favorite healthcare challenge that you think AI may have solved by then or a problem we won't have anymore?

>> Oh, that is a very good question. I think we'll have solved charting. So, technology is very much in the foreground of the clinical work we do today. I mean, you probably all experienced seeing a doctor and they're glued behind their keyboard. Everyone has to do a lot of charting. Our labor and delivery nurses have to chart every 15 minutes and even more frequently when they get closer to a delivery. So, that takes up a lot of head space, and we kind of take away our clinician's ability to stay in the moment and enjoy what they're doing. So, I hope we can automate a lot of that charting and kind of move technology from the foreground of healthcare and into the background so we don't have to think about it. So we don't have to interact with it, but it just takes care of the more repetitive tasks of their roles.

>> It really does. And lets doctors be doctors.

>> Yeah.

>> My mom was a chiropractor and I remember so many times watching her take notes and chart notes and wanting to play, and she's stuck doing that. I can't imagine.

>> They're glue to their keyboard. I'm sitting there like, "Okay. Look at me. Eyes on me. I'm the customer."

>> Hello. Hi.

>> "Am I okay?"

>> I'm a human. Yeah. Am I okay? Wow. Kris, this has been absolutely fantastic. Thank you so much for fitting us in this morning. I know it's a busy week for you.

>> Thank you guys for having me. It's been a pleasure.

>> Thank you very much.

>> Yeah, it's been absolutely fantastic. John, thank you for being here. Very fun. And thank all of you for tuning in to learn about these real-world applications of generative AI. We're here on day three of Dell Tech World in fabulous Las Vegas Nevada. My name's Savannah Peterson. You're watching theCUBE, the leading source for enterprise tech news.

Kris Kristinsdóttir | Northwestern Medicine

search

>> Good morning, AI fans and welcome back to fabulous Las Vegas Nevada. We're here on day three of Dell Tech World. It has been an awesome and insightful week. I feel smarter than when I woke up on Monday. My name's Savannah Peterson. You're watching theCUBE. Very delighted to be joined by John Furrier today. Good morning, John.

>> Great to see you. Great event, amazing content. The guests have been phenomenal. Content quality-

>> Yes....

>> has been amazing. Generative AI is here and we've got great use case examples and it's real.

>> Yes, and speaking of generative AI and real use cases, very excited to welcome to the stage Kristen . So, nice to have you here. You're a quantitative researcher at Northwestern Medicine. Welcome, first of all.

>> Thank you guys for having me.

>> How's your week been?

>> It's been great. I've already learned a lot. It's a great conference. It's my first time in Vegas.

>> Oh, my gosh.

>> So, quite overwhelming, but also very fun.

>> Oh, my gosh. Wow. First time in Vegas. John, do you remember your first time in Vegas?

>> Oh, God, no. I can't, actually. I've been here so many times. No, it's like, this is our home. We live here.

>> I was going to say.

>> We live in the Venetian. I could walk blind in my room. The tissues are over there. No, we love coming here. We love the stories. Dell's been great too. I mean, you're working with Dell Innovation Lab, right? HPC Innovation Lab?

>> We have. We've been working with them a lot. They've gotten some great help. They're a great partner in the sense where they help us kind of find the technology that we need. There's so many solutions out there, so they were able to help us kind of puzzle together what would work best for us and set up a great system.

>> Kris, I got to ask you, because one of the things we've been looking for, last year when we were here, we said next year is about workloads in production, real use cases that have real impact. Societal impact, human impact, business productivity, all across the board. This is this year. You guys are doing something really cool with generative AI that is not only cool technically, it's got training and inference, and you've got HPC in there. So, it's very nerdy, which is great. We'd love that. We love the nerd talk, but also you're doing something really compelling that saves lives too. So, I want you to share your story, tell us what you're working on and take us through that.

>> Absolutely, will do. So, I work with the research and development team at Northwestern Medicine. Now, Northwestern Medicine is a hospital system in the Chicagoland area. We have 11 hospitals and over 200 sites of care. And the research and development team is a group of 20 people who have backgrounds in engineering, data science, medicine, and nursing. And we all come together in our mission to try to advance technology within the healthcare system and improve patient care. So, we've worked with a lot of different areas. We've worked in women's health, we've worked with wound care nurses and radiology space. And our projects range from anything from being able to diagnose a disease to optimizing workflows and reducing physician burnout. So, one of our more exciting projects right now is actually a multimodal generative AI model that's able to look at an X-ray image and generate a report. And we have this model live in our hospital system. We have 50 radiologists using it since this summer, and we're seeing some incredible performance improvements. The radiologists are 40% more efficient, and they're also telling us, they're giving us feedback like, "Now we get to spend more time doing what we actually love to do."

>> Helping people.

>> Helping people.

>> Take us through the workflow and what the technology is. You've got data, you're scanning X-ray scans in, you're training it and you're making inference. Take us through some of the details under the hood.

>> Right. We actually have our model running on all x-rays that are taken within our hospital, that radiologists use a system to write the reports called PowerScribe. And in that system you have different templates based on your preference and our model, the ARIES system is just another template. So, the radiologists can choose to use this pre-generated report. They then look through it, make sure everything looks correct before they send it to the doctors and to the patient. So, that's how it looks like. We used to train everything on the cloud before. We drove all innovation on the cloud, and we still do for a lot of our projects. There's multiple use cases that the cloud works very well for. But what we were finding, we're finding ourselves as we grew and as we scaled up, we're finding ourselves more and more limited by our compute resources. Now the cloud is very rigid and also our data lives on-prem, so having to funnel all that data from our on-prem to a GPU cloud instance was becoming more and more of a bottleneck in our training. We got some incredible help from the Dell HPC Innovation Lab. They helped us kind of set up and puzzle together what would help us and what we now have are four PowerEdge XE9680 servers that are all on-prem and HIPAA-compliant. And we have eight Nvidia H100 GPUs per server. And we're seeing some incredible runtime improvements with this new setup. I can give you an example.

>> Please do.

>> Yeah, we'd love it.

>> When we first trained our model, this generative AI radiology model called ARIES, we use a sequence length of 40-96. So, a sequence length is the number of tokens that you can process at the same time. Now, when we trained this with a sequence length of 40-96, one EPOCH took eight days to train. What we've done now is we've scaled up to a sequence length of 200,000. So, by doing that, we're not only able to increase our resolution of our training data, we're also able to do training on much bigger studies. So, what we're doing now is we're expanding to CT.

>> Amazing.

>> Now, CT is 200 x-rays, images, and four different views, so there's so much more data. So, scaling this sequence length up is very helpful for our use cases, and we're able to do it in the same time as we did with the old ARIES model, the x-ray model. So using this new setup, we're able to go from a sequence like the 40-96 to 200,000, which means that we're doing 50 times more backward passes in our training, and we're able to train it in the same amount of time. So, we're in fact seeing a 50-fold increase in our runtime.

>> Oh, my gosh.

>> Which is incredibly helpful. And we're so grateful to the Dell AI team for helping us make this possible and make real impact.

>> Well, I mean Dell in medicine are also very synonymous. Michael Dell has a huge passion for that. He founded their own medical school. It's no surprise that Dell would be a partner wanting to help see this. So, now that you're able to do this better and faster, is Northwestern Medicine going to roll this out to other hospitals and other radiologists around the country, around the world?

>> It's definitely something we're thinking about, absolutely. We want to be able to help as many people as we can. But another vision we have, and another thing we would like to see is that every hospital has their own AI team and their own AI factory. I mean, the healthcare and life sciences industry is responsible for generating more than 30% of all data generated globally. And within our hospitals, we're only using 3% of that data. So, there's so much data that we're not leveraging that we can use to help us solve problems.

>> You know what's interesting, is that you think about the impact of getting that training set bigger and look at the different CT scans increases the aperture of what you can do. At the end of the day, you have customers, end users. They have to come in for an appointment and then a technician has to shoot the film, whatever the technical term is for taking the scan. They got to do the scan and then they got to process it, send that to someone to look at. The person goes home. Someone might not be on call, then they got to come back to do another angle maybe. So, in the workflow, there's a lot of inefficiencies.

>> Absolutely, absolutely.

>> And hassles for the customer, the person who might get overlooked.

>> Yeah.

>> So, I can see, how much do you guys put into that thought and can you share what the impact will be as this rolls out?

>> Absolutely. I can actually give you a good example. On top of this ARIES system, we recently added a feature that is able to prioritize the radiologist's cue to put more severe cases on top of the cue, so they're read quicker. And we just a week, I think, after we deployed this prioritization model, we had a patient come in on a Friday afternoon and ARIES was able to spot that this patient had a pneumothorax, which is a very severe condition, has to be treated immediately. Now, what sometimes happens on Fridays is that the radiology reads get carried on through the weekend. So, the radiologist actually won't look at it until the coming on Monday. But with our algorithm, we were able to prioritize that patient, make sure their scan got read by a radiologist, and that our clinical teams were able to intervene, which might not have happened if we didn't have it.

>> Amazing. And that has real tangible impact on that patient's life.

>> It does.

>> And also probably makes that radiologist feel better that it didn't go over the weekend.

>> Absolutely. Yeah.

>> The workflow of the practitioner, the radiologist, it's in the template.

>> Yeah.

>> So, you guys on the back end, so the back end, what's next? What's next with Dell? Because you get more horsepower, you have a lot of GPUs. Congratulations. That's tough to get those these days. It's always fun to get-

>> Yeah, that's a good point....

>> the GPUs, devours a lot of data. What's next?

>> That's a very good question. We have some very exciting projects in the pipelines. We're working a lot with text now and the medical record and being able to feed the whole medical record into a language model and do all sorts of different stuff. I think our long-term vision is, I mean, the healthcare system is very, very complex. At the core of it, we have the human body, which is by itself very intricate. We have more than 30 trillion cells and even more chemical reactions happening every second. Now at the hospital, we have all these sensors and lab tests and imaging studies that add to that complexity. And then running a whole hospital system and taking care of people is also super complex. You have all the billing and you have managing beds. We thought, what if we can model this? What if we can actually create some sort of digital twin for the system that will enable us to say, "Okay, if this part looks like this and we put it in this environment, what's the outcome going to be? What is the future of this patient? Can we predict future disease states? Can we predict response to treatment and provide a more personalized care?" For example, suggest an optimized diet. So, that's something we're very interested in.

>> You're pointing out a really good thing that we talk about a lot on theCUBE. When people think about multimodal foundation models, think about large language model, text, multiple languages, whatever, images videos. But healthcare is all about multimodal. Scans, DNA's a mode, you know, all kinds of things are now different modes and not text.

>> Right. -

>> Gen AI is a dream scenario for healthcare.

>> It is. It really is. I think healthcare is such a sweet spot for AI.

>> It really is.

>> And it's amazing.

>> You're excited, aren't you?

>> I am. I'm really excited and optimistic.

>> I'm excited now. I want to work there.

>> No, I think it's fantastic. So, when you were working with the team, how did you decide to tackle this problem first?

>> That's a very good question. We have an amazing director to our team, Dr. Mozzi Etemadi, who is a anesthesiologist himself, but also very knowledgeable about AI and generative AI, and he kind of helps us funnel the projects we want to tackle and think about what has the biggest impact. Another thing we are very focused on is getting clinical buy-in from the get-go. We don't go into a project without having a clinician champion to support us because it's very important that we're not creating something that we think will help, but then doesn't. That we actually work with our clinicians to kind of shape the future of AI.

>> That's awesome. Okay. Speaking of, now I'm curious because you're clearly very brilliant and doing some awesome work. What's your background? How did you end up doing this?

>> Oh, that's a very interesting question. I'm Icelandic. I'm born and raised in Iceland. I have a background in biomedical engineering, and after I graduated, I worked as a key account manager for a medical device distributor for a while where I got to spend a lot of time with clinicians discussing technology and integrating technology. I was so surprised by the state of technology within the hospital. After having learned about all the different applications of AI and healthcare and all this shiny technology, clinicians were just frustrated with technology. So, on my mission to try to get healthcare technology up to date, I moved here to the States, did my master's in bioengineering with an emphasis on AI in healthcare at UC Berkeley in California, and then got word of-

>> Go Bears.

>> Go Bears.

>> Absolutely Bears.

>> Yeah. Then made the move to Chicago. Got word of the research and development team. I've always been very interested in being able to work in-house alongside our clinicians to have the most impact.

>> Wow. What a journey.

>> Well, congratulations. You got a great mission.

>> How cool.

>> Love what you do.

>> Thank you so much.

>> We need more healthcare technology.

>> Is she our first Icelandic guest?

>> Yeah. I think.

>> I think she might be our first Icelandic guest.

>> Amazing.

>> You are representing Iceland.

>> Amazing.

>> Is there anyone back home you might want to say hi to in Iceland?

>> Absolutely. Everyone. My parents, my boyfriend. Yeah, everyone's home.

>> All right.

>> Oh, that's fantastic. Wow. Well, I really can't wait to see what you do next. I'm going to ask you one final question. Let's look way into the future. I mean, I love that you were thinking about AI and healthcare years ago, if this is what you were getting your education in, clearly someone with a lot of foresight. 20 years from now, what sort of healthcare challenges, and there's obviously a lot, but what's your favorite healthcare challenge that you think AI may have solved by then or a problem we won't have anymore?

>> Oh, that is a very good question. I think we'll have solved charting. So, technology is very much in the foreground of the clinical work we do today. I mean, you probably all experienced seeing a doctor and they're glued behind their keyboard. Everyone has to do a lot of charting. Our labor and delivery nurses have to chart every 15 minutes and even more frequently when they get closer to a delivery. So, that takes up a lot of head space, and we kind of take away our clinician's ability to stay in the moment and enjoy what they're doing. So, I hope we can automate a lot of that charting and kind of move technology from the foreground of healthcare and into the background so we don't have to think about it. So we don't have to interact with it, but it just takes care of the more repetitive tasks of their roles.

>> It really does. And lets doctors be doctors.

>> Yeah.

>> My mom was a chiropractor and I remember so many times watching her take notes and chart notes and wanting to play, and she's stuck doing that. I can't imagine.

>> They're glue to their keyboard. I'm sitting there like, "Okay. Look at me. Eyes on me. I'm the customer."

>> Hello. Hi.

>> "Am I okay?"

>> I'm a human. Yeah. Am I okay? Wow. Kris, this has been absolutely fantastic. Thank you so much for fitting us in this morning. I know it's a busy week for you.

>> Thank you guys for having me. It's been a pleasure.

>> Thank you very much.

>> Yeah, it's been absolutely fantastic. John, thank you for being here. Very fun. And thank all of you for tuning in to learn about these real-world applications of generative AI. We're here on day three of Dell Tech World in fabulous Las Vegas Nevada. My name's Savannah Peterson. You're watching theCUBE, the leading source for enterprise tech news.