Deal pipelines for healthcare

How to create outlook signature

hello everyone my name is Eduardo I'm a Solutions architect with Amazon web services today I have here with me Sanjay from IBM hello Sanjay hello uh hello my name is Sanjay dutt I'm a principal Solutions architect with IBM so we're going to talk today about IBM Cloud pack for data a lot of customers reach reach out to us and ask us you know how can they build new applications and new workloads with machine learning you know to build differentiation into their businesses right and improve the quality of the data and the results that they can get from this Sanjay do you have some stories that you can share with us today about how Cloud petrol data is helping some of these customers uh you know really differentiate their business yes so we have a reference architecture we have implemented at a customer and so we can talk about how this reference architecture helps predict patient outcomes also predicts uh readmission rates and you can build a lot of models once you have the data so let's take a look at this uh we'll draw the reference architecture stage by stage and we'll provide an end-to-end story so this is an implementation that you've done in club pack for data and AWS for a healthcare industry that's correct okay so here we are looking at the data sources that we have in this particular scenario right where you are collecting data from Health devices that's like patient vitals data right exactly so this could be like iot devices that are hooked up in the hospital or in the patient's home right so this could be a variety of healthcare device you know Health devices iot or some other traditional devices too and I see here that you're also then connect connecting to data repositories on an on-premises data center of some kind right exactly so this could be uh on the top uh the databases could be in the hospital the patient uh exactly yeah all right so so the way that uh I understand that we've discussed about this right so the the way that we are collecting this this information here if we think about the connected devices right they are pushing data into AWS and this data ends up in a Amazon Kinesis data fire hose and then from here we do a process to do data validation right so if you see any issues in the quality of the data that's being collected we are pushing we are using Amazon SNS to send notifications out into the an API right that you have running on on the on the on-premises side of the equation right on the clinics and the hospital to re-process this data right to send it across again and then I understand that once this data has been verified you know we are pushing it into Amazon S3 it is from where I believe Cloud pack for data is going to read this data to train the the the ml models that you're building correct right so this is uh this we will call this The Landing bucket or the uh raw bucket it passes for yeah data correct okay so once this data is there so uh that's where you know the cloud pack for data so it so uh so cloud cloud pack for data is a unified system for data and AI so you can take data and Infuse AI into it and it's made up of multiple components in it so let's talk about the first component in there so as your data is in the Raw bucket here uh this is uh Watson machine catalog it's an Enterprise data Discovery Tool uh it's also a metadata repository okay so who has access to what data and it also discovers this data so that will be helpful once what's once this data can be read by Watson machine catalog here then we can use data stage so data stage is an ETL tool so this both can work in concert the Watson machine catalog and our data stage so it can pull the data the right data and then you have developers here they can build models using IBM Watson studio so these are AI models those are built and once those models are built they can be deployed using Watson Studio into Watson machine learning language okay uh and so the Watson machine learning will deploy those models and it will generate these endpoints those endpoints will be then used by these Healthcare applications to make critical predictions right okay so these are inference endpoints that Healthcare applications using to predict the patient outcome exactly and so though Healthcare users use this Healthcare application which will provide those predictions then I believe that you are combining into this equation here then with Watson knowledge catalog the data that's coming from the the patient systems of records that are either on traditional databases or traditional applications right right and you're combining this then the iot data with the patient data the patient history data to generate this this predictions that's exactly right okay and so the data stage here is is being used as an ETL tool to do data preparation to train the machine learning models right exactly so it will do some basic uh formatting data prep uh but the Watson machine catalog will also can apply rules as far as like data masking okay who has access right okay yeah because we're dealing with sensitive patient information right exactly so the data once data stage is prepared then it also pushes into S3 right so this would be something like the prepared data for our machine learning uh training right exactly so this this is like the curated data once those machine uh data stage has read you've created these models so this particular S3 bucket uh could be your curated data that's available for any of these applications to then read okay and how this compact for data uh acts as data repositor is on AWS it does it have some kind of connectors that allow users to to do that yes so they are pre-built connectors to uh IBM's uh AWS data sources and also to third-party sources okay now because we're talking about uh about dealing with uh you know personal identifiable information and sensitive patient information I understand and on this particular use case you're also using another IBM solution to make sure that the environment is secure and there is no no threats or any anything like that right so we want to talk a little bit about how have done that yeah so you're right you know security is extremely important and also we want to figure out if there are any patterns right uh which are very varying from the normal so there are uh we have ibmq radar uh it's a security product and basically it analyzes data from the VPC flow logs here and also from AWS cloudtrail okay it reads those uh logs and figures out there are Trends if they vary from the normal it will flag let the security operators know so this would of this flow would be something like this right where we have the uh these logs being analyzed by Q radar right and then if you find anything out of the ordinary if you find anything that's important or you know if you you generate an alert for security operators then yes exactly this so the security operators they have access to the uh send curator Sim and curate our Sim is it can you can build predictive models using that uh it does proactive monitoring uh so it just make sure that your environment is secure and anything that uh is out of the normal will get flagged okay so this is a very interesting implementation and you have this already running in production with customers yes customers and AWS customers as well that's right okay and I see here that the cloud pack for data is running on openshift So based on what I see here you know we can deploy Cloud pack for data on on Rosa on red hat openshift service on AWS that's correct how does that help customers so Rosa is a managed offering so a customer doesn't have to worry about patching or maintenance any kind of regular day-to-day so they can offload that uh to the Rosa service and focus on creating strategic Innovative products all right yes all right this is very very interesting thank you Sanjay really like talking to you today and thank you everybody we'll see you next time thanks everyone bye