Hello everyone. It's such a pleasure to be here with you all of you today. So my name is Rinni Vasa Ti and I currently serve as a engineer lead at one of the largest healthcare company in United States. Let me talk briefly about my experience. So over the past several years I have worked at the intersection of data analytics, healthcare systems, and automation. So today I'm truly excited to share how advanced analytics and artificial intelligence or transforming clinical robotics and the future of healthcare. In this session we will walk through how data and a are shaping healthcare's next frontier from from precision diagnosis and robotic surgery to ethics, compliance, and patient centered innovation. By the end I hope you all see that e and robotics are not just technologies but they are instruments of compassion, precision when built and trust and data integrity. Let's move forward and see how it works. So let me move to second slide. So let's talk about a, let's talk a little bit about abstract and overview. So let's begin with a quick overview of what's that? What's this transformation look like? So healthcare today is moving rapidly from reactive to prior to models of care, fuel by data analytics and a power insights. So we call this as a data driven evaluation of healthcare. Imagine a system where every heartbeat, scan and record isn't just stored, but learn from the real time. So our unified framework brings together SaaS analytics, mission learning, and robotics systems. So forming a single ecosystem like that, that suppose predictive maintenance, precise diagnostics, and continuous improvements. But this evolution also requires a strong ethical foundation as well, right? So we have to ensure that all the government regulations and, P-H-I-P-I data security, and then HIPAA and GDP are compliance, transparency, and explainability because healthcare because in a healthcare trust is the ultimate currency, right? Yeah. So that's what we are going to talk about today. So let's move on to next one. The healthcare transformation. So let's talk about the larger transformation happening in healthcare currently. So we are witnessing a paradigm shift from treating specific illness too, predicting and then preventing it. So this shift is powered by a data explosion, right? So now we now collect the data from multiple sources from amazing. Genomics and then wearables, the wearable devices, and then IOT devices and electronic health records, and an incredible multimodal data set that can completely redefine how how how we understand the health overall. And but here is a challenge that the integration is very challenging here because all of this data lives in a different systems that don't always talk each other. So connecting clinic, clinical robotics with AI analytics, but requires a robust interoperability framework. That means we have to make a. Seamless channel or connection between all these sources to talk each other and bring the data with the, in a near real time or real time. So that, the clinical robotics will take a decision based on the the collective data from all the sources. So when we achieve this when we achieve that connection, we can reduce medical error. And then accelerate diagnosis and enhance both patient outcomes and overall quality care as well. So let's let's move on to how the, actually the clinical robotics evolution happened. Let's step back then, let's let's talk about let's step back in time and see how clinical robotics has evolved In the 1980s and 1990s systems like the Puma five 60 and later, the Daven c surgical system transformed minimally in VA surgery. They showed us that emissions could enhance a human precision in the, in 2010. Robotics center, diagnostics and rehabilitation, helping stroke patients regain mobility and supporting the re procedures with a high year accuracy. So by 2020, AI driven, robotics begin, stepping in nearly every domain like a surgery, pharmacy, automation, elderly care, and hospital logistics. Now we are entering, in a era called, I call it a integration era. So where robotics data connects directly with the patient analytics to enable closer loop care means like a system where feedback and action continuously improve outcomes. That's, that, that's important here. When it comes to the integration area. So let's move back. Sorry. Let's for move forward and see how AI and mission learning in healthcare really, play a works. So by touching all the key important points, like a deep planning models, how the deep planning models work and neural language, sorry, natural language processing and the predictive analytics. And then, reinforcement learning, right? And then finally, now the precision medicine. So let's talk about each concept in a very high level. So deep learning models can now detect disease and CT scans. MRIs and pathology slides faster than faster and more accurately than ever before. Right? And natural language processing. So our visualization, NLP usually allow us to extract insights from doctor notes and unstructured text turning free flowing narratives into structural data. Then let's, and then move to predictive model. Predictive analytics helps us detect sepsis, heart failures or infections before they come. Become a critical it's about moving from cure to prevention. So then let's talk about the reinforcement learning. So it, it is being applied to a robotic surgery teaching robots to adjust their movement. Based on the outcomes, much like a human lens through the feedback. And then we have a precision medicines where ai analyze all the genetic genetic data to design a personalized therapies for each patient based on its condition and many other factors specific to that particular patient. And then condition. So this is how we move forward, truly individualized healthcare. Yeah, so I like to say here in the past, doctors treated patients in the future, the data will help doctors know their patients well before they even arrive to the, visit. Yeah, that, that's all that's all about the AI mission learning and healthcare. In this in this slide, and then let's move forward to the evolution of healthcare analytics framework. Has robotics and ai matured? Our analytics frameworks evolved alongside them. So like we started with a, descriptive analytics, simply reporting what happened. Basic descriptive analytics is nothing but take all the existing un historical data and what we have. In what our databases or in our systems, and then just simply reporting what happened, right? Then predict two analytics came into the picture. So predictor analytics means like applying some statistical models on top of what on the top of the data that we have right now in our system, and apply some mission learning algorithms to forecast outcomes. So now we are entering into the era of prescriptive analytics. Descriptive is done. Analytics, predictive is done. And then now prescriptive analytics means recommending what actions to take next based on predictive insights that we have, predictive analytics, and we have insights from that. And then now. Use this, by using this ai LLMs you can just based on the predictor insights, you can just recommending what kind of, actions and, course of treatment that we can design for the specific patients and specific, conditions. This shift is covered by a realtime data pipelines, like using tools like a Kafka spare and then Hadoop alongside advanced modeling platform such as vo right. And together all these tools enable continuous intelligence where analytics doesn't just analyze, but it acts on the facts of the data. And when we visualize in dashboards and KPIs, key performance indicators, these insights empower clinic clinicians directly bringing data from the server room. Server room means the data from all the source systems to the bedside means where actually patient and and the doctor actually, treats the patient. So he has everything right now, next to the patient and the next to the bed. So the doctor's nurse, you don't have to really go research, by diving into multiple data sources and then, take a call based on that or data. But that AI and the system will really do it for the doctor so that they can just quickly take a call and take a decision and then, by providing some prescriptive NLA, on treatment. So let's move on to. So proposed framework. So what we are trying to propose here, so let's talk about the five layer architecture that brings all these components together. So we are divided into different layers, like a data equation layer, processing layer, yeah. And analytics layer, and then integration layer, and then presentation. So let's talk about each. So data equation layer collects inputs from robotics or electronic health records like EHS and iot sensors, devices, and emerging systems and all other devices, what are all medical devices Processing layer. So processing layer converts this raw data, whatever we getting from all the devices and what we. In this processing layer is basically like we consider the data from all these, systems and then we convert this raw data into structure and usable information. Excuse me. And then the next layer is like AI and analytics layer. So where mission learning and predictive models turn the data into inside, right? We have a structured and usable information from the presentation layer. Now we apply the all these AM models in this analytical layer and where mission learning and predictive models. The statistical models turn data into insights. They give, oh yeah this is what the data shows tells us. And then this is the information. You have it. And then, then but still, right? Those are like a no independent data points. Now integration comes into the picture. That's why we call as a integration layer. So APIs and middleware, right? And application interfaces and the middleware ensures that interoperability across the systems we have all the data points, data, source data objects, right? And then you need to integrate. All the, data from different systems like I sensor systems, scans and everything. And then the move, the data has to be collaborate. Move between each systems and then, collectively give make some kind of summarized insights. With that data that's called integration layer. Now presentation layer. So when we have all the data yeah, integrated data and provides provides the dashboards alerts and patient support directly to the clinicians, right? And together these layers form a unified ecosystem creating end-to-end automation and delivering insights the at the point of care. So let's move on to let's move on to the data pipeline architecture. So this slide this slide shows how. How that architecture looks in motion. And first Apache Kaka, we, let's talk about some technical technical components here. First Apache Kafka streams, robotic telemetry and patient vital in real time. Apache Kafka, really streams, flows, sensor data almost in real time. Like a patient vitals or robotic er, whatever, robotic tools actually, performing at the point of time. And then it talks to the patient vitals at the same time, and then Apache Kafka. So process these data streams across the distributor cluster. And then usually we have a huge amount of data so for faster processing or accessing, we, dis, distributed into different clusters for the parallel processing. So these Apache spa process, all these, cluster data. Very fast. And then next to Hadoop, HDFS, actually Hadoop, HD FS is the file system storage. Excuse me, the structured and unstructured data securely. So HDFS has the capability of storage, like even structured, unstructured, and, we can use we can implement proper governance on those data elements. Now let's talk about the SaaS wire. SaaS wire. It's it's an analytical tool, right? I mean it's it adds analytical power power to that data which resides on the HDFS and then visualization forecasting and predictive modeling. And finally, HHL seven and FHIS standards ensures interoperability with hospital EHR system, right? That's is interoperability like. Communication between all these different data points. They ask for health electronic health record system. So together, together this pipeline entire, the pipeline allows a low latency decision making in robotics workflow. So meaning. Meaning like a clinical rewards can respond to changing conditions within a millisecond, right? When we read the data in real time, when we stream the data in real time, when we have the data, updating, almost in a. In real time. So robotics should be able to capture those updated, vital or, whatever that data patient data that come flows through and should take an action should, guide the patient doctors almost like a, instantaneously. So that's, that, that's what another final goal is here. So with another, the robots should respond in milliseconds. In essence, this is the digital nervous system of a modern healthcare. Okay, let's talk about mission learning infrastructure, right? When to make AI truly operational. We need the right infrastructure now. Yeah, we talked about different, software components and, different pieces of technology, how it works and how it really integrates the data and it provides the interoperability interoperability feature. But in order to make you know all this to work, we need the right infrastructure. That's where the automated ML lifecycle covers automated Emission Learning lifecycle basically covers the data preparation, featured selection, and then date model training and validation, all seamlessly automated to save time and without any bias. And a hybrid cloud setup it's like a hybrid, like you have an on-prem. You have a better control over all the go, data and governance applying some securities and, control access to specific PHI and PI data to the limited, very secure people, secure, I mean in a secure way. And, and then when it comes to the cloud the cloud gives you the opportunity either, what you call them mean cloud actually help you to scale the product ally or vertically, based on your need, based on your data, based on your processing power, you can scale up and scale down. So there a s and does, it comes into the picture. So basically like an hybrid architecture, you can secure the data. You can, scale the scale, the product. That's how a we hybrid cloud setup and obviously GPU cluster, right? The GPU clusters enable, deep learning models to understand the images and medical images like container medical images. It requires a lot of, computational power. So here the GPUs actually works very well. For the, to learn, to understand and, know deep, take a deep dive into the medically amazing things. And then now container is deployment. Container deployment is nothing but irrespective of the operating system, right? And underlying operating system, the tool and the product should work seamlessly because they're like nowadays the, there. Infrastructure and underlying operating systems are sh should not really, limit the CAPAs limit the capability of the product that you're building. So that here, the containerized and deployment ensures the flexibility across multiple, across environments. Now models are continuously monitored for a drift. Models are continuously monitored for drift fairness and performance. Okay. And then, and when Drift is detected retraining pipelines automatically kicks in. So some of the key algorithms in play right include CNNs, like a convulsion net natural networks sorry, convulsion neural network for image recognition and recurrent neural networks, RN for patient time series prediction. And the transformers for clinical text. And then reinforcement learning for robotics control. This synergy enables the real time adaptability where robots and algorithms keep learning, not just, working and that they should, train themselves. They should keep learning and then in keep working for us. Okay, so let's move on to data governance compliance. Of course. With all these intelligence comes responsibility and so in healthcare, data governance and compliance are non-negotiable at any cost. Very important. And that our systems implement HIPAA controls including role-based access like a ES 2 56 encryption algorithm and full audit logging. And we align with JDPR principles such as consent tracking and privacy by design. And of course, line know security architecture involves multifactor authentication and multifactor authentication and network segmentation and what we call that now, intrusion detection. And we maintain a data integrity through metadata cataloging and linear tracking and data quality qua scoring. Ultimately this builds a chain of trust. Ensure like every data point is traceable and then ethical and then transparent. Makes sense, right? And as often tell my team that in healthcare, like data doesn't just need to be accurate, it needs to be accountable as well. That is most important. Hope I understand that. Yeah. So let's talk about next clinical applications. So you know, some practical application, look at some practical applications right in He enhanced AI enhanced robotic surgery. The workflow is divided into three stages, like a pre operator, in operator and post operator. So ai creates a three DR one maps that helps surgeons plan every incision with a millimeter position, right? That is very important because some of the organs are very sensitive and very minor, so it has to be the incision. It has to be very accurate. And then intraoperative. Real time tissue classification guides decision during surgery. When the surgery happening, right? The real time tissue classification when you. When you put some incision on any of the tissue and that that, that particular tissue need to be the robo tick the component that, that is doing the operation the robotic operation, that has to read, understand the understand the tissue classification. What this tissue is and how it reacts when we do an incision. And then how it actually, works and how it actually bleeds and everything you need to, the robotic that should robotic tool should understand this tissue classification and then guides the decisions during the surgery. So it has to happen always like a human right When doctor, when surgeon cuts a tissue. He knows what that tissue is and how it works. And then, if something, she finds something unusual, he should take a take call here and he should take a decision and, he has to change the course of surgery or maybe, do something. So in the same way, like we expect the robotics, should do here in this situation as well. Then postop. Predictive models actually track recovery and identity communications early complications early. So this predictor, we do right, predictive means like applying some statistical models on top of what we know, what we have right now. And then basically it tracks it, track the recovery and identify the complications like. Early. Okay. Yeah. This surgery this surgery gives more complications in this particular area. So we cannot simply guess it, right? And take the data based on historical surgeries, similar kind of surgery, what happened, and and then what are the complications and, and then apply some static statistical models and that data. And then, and then compare it with the current surgery. And then, then comes up with some kind of, recommendations and saying that, hey, this surgery or this post operation, these are the, these probably these are the, complications may arise and then, guide the doctor and then treatment the patients through doctor that, the post operator. So these innovations have reduced to operator times by 23%. And then complications by almost, one third like, and then beyond the surgery, right? And he also supports predictive maintenance forecasting robotic component failures with the, over 90% of accuracy and reducing downtime by nearly 40%. And finally, now, patient risk satisfaction, right? Satisfaction. Patient stratification, like where digital time as scoring helps hospitals to prioritize, high risk patients to and allocate resources smartly. But that, that, that's yeah. So that, that, that's a key here, right? That, that's it. This helps total hospital admin, administration and to control the resource wastage. And also right person, right treatment the critical person gets a treatment for us. So each of these use cases demonstrates that when analytics meets robotics. We don't just treat patients faster, but we, we treat them better. Let's move on to, let's move on to emerging technologies. The next frontier lies in four kind of technologies, for for powerful technologies, quantum computing. Edge computing, fer learning, and finally explainable ai. Right? And then so let's talk about all these all these that powerful technologies, quantum computing will revelation, revolutionize the genomics, drug discovery, and molecular simulations by processing massive dataset at quantum speed. Edge computing will push AI closer to where the care happens besides bedside phases or even insert surgical robotics, really surgical robotics, reducing latency, and improving privacy. Now federated federated learning. So we enable hospitals across the globe. To train a models together without sharing sensitive data. So a perfect balance of, collaboration and privacy. And finally, explainable tools like, SHA and know Lyme health clinicians understand why and made a particular decision building. Trust and transparency, right? The clinician, that's the most important thing. They should understand why this particular large language model or some kind of together them really take, made this kind of particular decision. So that I know you can just try to, build that trust, build that model output, and also you can in a better position or surgeon in a better position to explain to the patient why they actually, that, that kind of decision has been taken here and these emerging technologies or extending scalability, collaboration, and plus across the entire healthcare ecosystem. Let's move on to challenges and limitations, right? So now it's very important to be honest with you about the challenges what we face here. So from technical standpoint inconsistent data quality, interoperability gaps, and high computational cost remains bad. Yes, no doubt about it. Because the data comes like a, anything now. So many variables, so many iot devices, so many, systems. Every system sends the data in its own way. And decoding the data and, understanding and then making more, usable data. It is very challenging. And that is called, interoperability gaps, right? You get this data from here, you get this data from, you get this data from here, and then you need to you need to make sure that you know all the data data points are. Should be should be, understand in a way that makes sense to other systems. Bringing the right point of data from each system and, understand the data in a, meaningful way. And in order to do that, like it requests a lot of computational power, computational cost. So means lot of gpu, a lot of, kind of CPU power and lot of, things like that. Cause a bit of more comparatively. Keeping all the systems separately, right? That it has advantage, but it has, some cost involved. So that is one challenge. And the organizationally we face clinicians' resistance often. That, that makes sense too, because like you. Human no one. As at this point, I don't think any mission can really be the human brain in, when it comes to take a decision. And then when you call when do something that is, you humans can feel the thing, not the admissions, right? I mean that, so that clinicians. Can resist to, accept some changes and often due to workflow changes. And then we need to widespread upscaling. That, that requires some kind of, training and adaptability is required. And then there are ethical concerns now, algorithm bias, black box decisions, and unclear accountability in AI assisted outcomes. It's always there. Every new thing, new technology always have there you some kind of know concerns. And finally we have a regulatory and clinical validation requirements. Models must provide themselves not just in theory, right? And, but in real world evidence. So it has to, prove that, okay, this is not, I'm just, telling you or giving you the results, not just based on the input that you provide. Now, these are the no real time scenarios happen, different, cases, different patients and the different, situations so that makes actually, more convincible and then more meaningful. So while the technology races ahead. Our responsibility is to ensure empathy oversight and human. Judgment always remains at the center. Okay? And then because no matter how smart AI becomes, it must never lose a human touch. This is very important. So the, let's move on to best practices and recommendations. How do we move forward responsibly, right? And start with a small pilot project always as we do and validate a clinically. Yeah. And then scale gradually. So that's the way we do projects are, the things everywhere, right? Rather than just jump and do it, large scale and, for all third user bases, we should slowly start small and then go large and then adapt. All the standards like h Hhl seven FHAR, and then API the first architectures for interoperability and use hybrid cloud models for flexibility and compliance as we talked in the previously. Cloud and, hybrid cloud is very important here. And then implement continuous monitoring for fairness and bias, right? Create ethics committees and, and data steward stewardship roles to ensure the accountability and most importantly keep clinicians, engineers, and patients in the same conversation. Because if I talk something, I mean as an engineer, if I talk something, clinicians may not understand and what clinicians talk, I cannot understand fully. And patient, he cannot understand what we are talking, and it has to be very the more. Layman kind of conversation when it comes to now patient, right? You should understand and everyone should understand like what we're talking about because that is a collaboration must is a real backbone for this innovation. Okay. And and then let's talk about conclusion and future vision. Okay. As we look ahead I want to leave you with this vision. So the convergence, the convergence of air robotics, AI robotics and analytics is the most transformative trial triad in modern healthcare. So we are entering an era of predictive, personalized, and. And participatory care. So where the medicine becomes proactive, not reactive, the future includes quantum AI for genomic discovery, edge robotics for remote telemedicine, right? And federated ai like for global collaboration, right? And then, but above all, it's about keeping healthcare patient centered guided by empathy, transparency and ethical responsibility. So let's, so let's build a system that just, that don't just, process the data, but truly understand the people behind it. And as I like to say, the stethoscope defined in 19th century medicine. Data and the AI will define the 21st century. Thank you all. Thank you all so much for listening, and I look forward to, c other, presentation in this conference. Thank you so much and have a great. Rest of the, we can, we, present our conference time. Thank you.