Real-time Generative AI Solution Architect | Roles & Responsibilities | Focus Areas

Hello friends how are you doing today so in this video we are going to talk about generative a solution architect roles and responsibilities and what exactly the generative a solution architect does in the day-to-day life okay so when it comes to the generative AI the world is going towards generative Aid Technologies and you can see let lot of Enterprises developing the gender da models and training the model and releasing it to the public and you can use those models right and also other companies like the domain specific companies industry specific companies are interested to use those generator models to solve the major critical business problems right uh keeping this aside I just wanted to touch upon this solution architect role what exactly it means for gender TV tools and Technologies so when it comes to the solution architect it's a broader category everyone knows that every stream has their own solution arcade for example if you are working for a platform platform work in an organization you will be called platform solution market and if you are doing some type of infrastructure setup infrastructure architecture all those work you are called infrastructure solution or debt and similarly if you are doing some type of application development work application design and application specific work then you are called application solution or critical the list Will Go On Right similarly when it comes to the generative AI solution noted the generative solution is not it has the knowledge experience and learned about the tools and Technologies which are useful for an organization to achieve this a Solutions so if a person is doing this type of works and if he has that background with a data engineer data scientist and those are the profiles or candidates will go into this generative a solution architect role when it comes to implementing a generative AI solution for any organization what exactly this generate UI solution not that does whether they are doing the regular work of to generate a solution or that who are doing the platform work infrastructure work or application work or how they differ how their work differ from the regular solution object that we have seen so far so this generator a is only snorkel has been emerging and if you evolved with the background of data scientist a data engineer you are the right person to pursue this particular solution or that in generative field I would like to share this chart it will definitely be useful who are aspiring to become a generative a solution Market let's go so I have put together a couple of charts to explain about uh the roles and responsibility of generative a solution architect and what exactly they do in the day-to-day life and what are all the areas they have to focus to improve their generation is not the career so these are all the rules generative UI solution Market performs in an organization so when it comes to an organization you must have a business use case to implement with generative a solution right so for that you must as a generation solution Market you must talk to the business team to understand right business use case to be implemented with generative AI solution okay and on sort of identifying the business use case as a generative a solution is not that he or she works with uh different teams to identify the right model to implement basically there are certain models available as a pre-trained model you can pick that and use it or else you can create your own model and use uh for addressing the business use case problems and if you are using any pre-trained generated AI models uh just uh make sure that you get to know more about the pre-trained uh General UA models like Foundation models or Gan or vae models all those sorts Etc and in case if you are creating your own generated AI model you must work with uh data scientist and AML Engineers to create the model as a generative solution Arctic and as a generative solution not that you must also do some research on the algorithms that's going to be used in your own generative a models also if you are using pre-trained generator models get to know about what are all the algorithms used in the behind of this pre-trained generating a models and next thing is as a generative a solution or that you must know about what type of platform required for building training and deploying your generator a models and these platforms might be like a VMware imagine or the container or any cloud service providers and also you must have a good understanding of deep learning machine learning NLP and nlg Frameworks like Keras tensorflow or pytharch Etc so these knowledge is very much required for that generate is a solution Arctic and as a generative solution or that a solution orbit you must have good knowledge and experience is not Transformers like libraries used in the Frameworks like a hugging phase or genism or core NLP these are all the Frameworks available and you must know about the libraries being used in this Frameworks and after building training and applying the model you must know how to validate the model because validation of the model will prove that whether you develop the model in associated with the business use kit that your organization came up with right and it has to be validated thoroughly to address your business problem them and when it comes to the larger models being used in any kind of Enterprises you must know about the performance requirements how you can maintain the uptime and what type of algorithms being used and what are the what are all the quality of that model being developed all those terms as a generative resolution Market you must know of and also coming to the cloud era right A Cloud area where you want to deploy your generator a model in either IBM Cloud using Watson or AWS or Azure cloud or Google Cloud you must have a clear and thorough knowledge and experience of those generating a models being produced by these cloud service providers so these are all the roles of a generative a solution are dead being performed in any organization so the next slide we are going to see The Works being done by this generative a solution architect so basically when it comes to implementing a solution for any business problem this generative a solution arcade first starts with identifying the model record for addressing this business early business model right the model can be pre-trained model as we discussed in the previous chart or the data engineer or a ml engineer or data is we create the generator model using python code so I took that example here the solution arcted codes and creates this model and he or she does this training of the model and once after after training he can prepare for deployment so the preparation for the deployment is lot of steps first after creating the model and after training the model you have to export the model to the deployment platform and after exploiting the model you can choose the right deployment platform and set up the platform and load the trained model into that deployment platform so here in this case you can take an example of creating a containerized deployment environment for your generative a model to run so for example you can use kubernetes or redial openshift or any built-in platform provided by the cloud service providers like IBM AWS as your Google cloud and they do have the platform available for deploying your trained model so you can load the trained model and create inference pipeline basically the inference pipeline is required for getting the data from input source and do a pre-processing of the data and post processing of the data are being done by this inference Pipeline and once after creating that you will use those data to train the model and then if you are going to a expose your generating a model as a service you can create a AP endpoint and expose the 10 point to the clients who are going to use your model and these endpoints maybe your rest AP endpoint which are going to be called your generative a model so these are all endpoint URL being used to call your generated a model and next one is monitoring and scalability so when it comes to that the large data set model right in any interface if you take they use a very big model that will be having lot of resources that will consume a lot of resources in your platform and infrastructure for that you must know about what kind of monitoring tool we use to monitor these type of large data set models and when it comes to the security right once after monitoring and scalability requirements are implemented for your generate AI model running in your platform either in the cloud or on premises or containers wherever and you must focus on the security so normally when it comes to the generator AI model there are advanced serial attacks happen that means like when your model is running in any environment there are false attacks come to your model and then it will collapse the model so for that you must focus on adversial attacks and Implement those security requirements for safeguarding your creative way model and also next one is after implementing all these things as a generator solution architect you must focus on testing and validation you must have the separate dedicated team for doing this kind of sub testing and validation because those teams are as expert in inputting the right data and prompting the model and validate the model and after doing all these steps you must focus on documenting this generate a model so basically this geneti model being used in your Enterprise suppose if you are developing it for the public you will expose it as a public generative a model in that case you must have a detailed documentation that will help users to use your model effectively and when it comes to Integra rating with other tools and Technologies other applications running your Enterprise you must know how to integrate using cacd Pipeline with your generally a models and finally you must know how to maintain and what type of version update if the algorithms are being used from different framework how frequent you have to take that version and update and maintain so this is how gender da solution is not that work in any Enterprise to create and use this generative a model either for Windows purpose or for the external usage and coming to the focus areas of generate AI solution or credit as a generating solution organic what are all the focus areas you must be aware about first one is deep learning fundamentals so normally when it comes to generating a model creation and if you are a solution market for generative models you must know the NLP energy algorithms and deep learning fundamentals all those terms and also when it comes to the standard UI solution or gets role as a solution arcade you must care about the scalability availability reliability of your generator a model if it is going to run in your in on-premise infrastructure or either public cloud or private Cloud wherever so that's a generator solution is not that you must have the knowledge about all the non-virtual requirement areas and deployment and productionization when it comes to generate a models when you create and deploy it in your on-premises or production environment wherever it is going to run you must choose the right platform right infrastructure and the right area to deploy your generated a models and also model architectures there are a lot of model or Pages available like foundational artificial model and Gan arbitral model via architectural model all those architecture models you must aware as a generative resolution arcade and then hyper parameter tuning when it comes to uh tune fine tune your generative a model whatever you developed or you occurred it at a pre-trained model you must know how to fine tune that model with the right quality of data so that your generator a model can serve better for all the business purposes and loss motions when it comes to loss functions there are certain logs persons like activation function loss functions all those are available in the library of any framework you must know what loss function needs to be used for developing your generative a model and also Evolution metrics so when it comes to Evolution Matrix the time the response time like Inception time Inception value so those type of metrics you must aware about it and also you must track those metrics while testing your model and also while implementing this model for any of the business purposes and also transfer learning you must know about what are all the Transformer models available what are all the Transformer techniques available for implementing in the generality a model that you are going to create and also when it comes to Resource Management if you are putting all these models in any infrastructure how to manage the resources what type of virtual machines needed what type of CPU required and what type of GPU required and why you need to go for a high storage mechanism all those things you must aware about um if you are going to work as a generally very solution market and user experience so once after deploying this model as a public model available for anyone to use you may need to put together a UI required for generating this connecting this generative a model with a nice UI or uxr and also when it comes to data preparation and processing you must know what type of data required for your generating a model to be trained and used effectively in your business organization sometimes the data may be available in your organization a small set of data or you may need to get large data set to train your entity model that the large data set can sit in your database or you can extract from any other data sources and when it comes to security you must know about what are all the security mechanisms regulations and governance all these things for your generator a model to run finding area of the environment so basically when you want developing energy model and you are putting it in your in any of the infrastructure the security regulation complaints are key things for example if you are generating a creating a generative a model for medical usage right so the medical usage must follow this HIPAA compliance right so all those steps should be um considered while implementing the generator model for your organization and also as a general area solution argument you must focus on use case and domains basically what type of use cases required for your organization or if you are developing a generator model any or any public usage for example the finance industry or insurance industry or any type of medical Industries all those domain specific knowledge you must have and keep learning about the generative a advancement new features go to the conferences and attend webinars all those steps you must do to become a successful solution Arctic and also communication collaboration this is key part for every solution or click to know of right in your organization you may need to deal with your business people technology people and ctOS so you must have the key communication and collaboration skills and regulated compliances as I said when you deal with generating a nowadays the complaints are key things because the data being used for training engine Ada model or the sensitive data and you must follow all the regulatory compliances while using the data for training your generation model and during the development of generator a model or during the deployment of gender Da Model there are definitely the problems um will arrive and so in that case you must have the hands-on experience in problem solving and troubleshooting field you must know how to debug the flow how to implement the fix for any problems all those steps and finally when it comes to the documentation it is key for any generator resolution or kit to document generative a model usage and best practices followed during the development and deployment so that's all about the focus area of a generator isolation Arctic I'm going to share this chart to you and keep focusing on these areas and get to know more about new advancement coming in general Tai and try helping your organization to solve all the business use cases but also when it comes to exploring this 1080 AI there are plenty of materials available in IBM Watson site and also AWS Azure and Google Cloud site you can go and explore about the new features coming in those areas if you like this video give a thumbs up and share it to your friends so that they will also get benefit out of it take care bye

Channel: Architect IT Cloud

Views: 5,049

Keywords: Realtime Generative Solution Architect Roles and Responsibilities, Model creation, Train the model, Focus areas of Generative AI solution architect, Scalability of Generative AI models, Resource management of generative ai model, Exposing generative ai as service, Generative ai platform, on-premises, VM machines, AWS, Azure, IBM, Google Cloud platforms, Generative AI Security, Governances of Generative AI model, Free material, Free presentation, Interview

Id: A-DOFuOlMBw

Length: 19min 7sec (1147 seconds)

Published: Sat Sep 23 2023