LTIMindtree Data & AI KenAI - 6 Weeks PoC

Introducing, LTIMindtree KenAI, an integrated mindful automation framework to help accelerate your MLOps journey on Azure cloud platform. KenAI, helps to standardize and streamline the AI/ML journey by leveraging it’s built-in components which help to operationalize the models with speed and scalability.

Leveraging the Azure native services, KenAI helps data scientists and ML engineers to accelerate the model operationalization process. It enables risk officers with a quick snapshot of model quality and provides business users with model explanations and what-if analysis. It helps optimize collaboration between multiple teams across the organization, by providing observability across the models. It provides insights around drift management, ground truth evaluation and model explanations along with what-if analysis. The outcome is full visibility on model performance, easing the process of managing any misbehavior in the models.

Some of the key Azure components included as part of KenAI are: Azure DevOps | Azure ML | Azure Kubernetes | Azure Blob Storage | Azure Insights | Azure Monitor

The key components of KenAI to help scale the AI/ML projects on Azure data platform includes:

• KenAI Automate: Provides automated pipelines for training, inference, and drift along with reusable components for model lifecycle management
• KenAI Assure: Provides model quality assurance with in-built test cases to help quantify the model quality and performance
• KenAI Govern: Provides comprehensive model explanations and governance on serialized models to drive transparency
• KenAI Monitor: Provides unified monitoring of the models deployed to ensure optimum performance continuously.

Note: The Scope of PoC will include operationalizing one standard machine learning model, along with showcasing model quality assurance, model explainability and model monitoring. To showcase model drift, it requires the availability of ground truth or baseline dataset. The PoC is about model operationalization, it excludes model development and any form of model-building capability. Our 6-week POC engagement plan includes end-to-end MLOps for one machine learning model leveraging KenAI as explained below:

Week 1: 

• Understand the current Azure Machine learning environment along with reviewing the types of models and datasets which are in production
• Finalizing the model and the artifacts which will be part of POC
• Setting up Metadata stores, covering key aspects like database and schema for the selected model and datasets.

Week 2 and 3:

• Setting up KenAI Azure Services like Virtual Network, Virtual Machine, SQL Server, Key Vault store, Blob Storage and Azure ML workspaces
• Migration of existing userbase and databases to Azure SQL Server  
• Migration of existing codebase to Azure DevOps Repo
• Deployment of KenAI on Azure VM once the migration is completed    
• Modifying/Updating KPIs for the selected model to showcase the model performance

Week 4, 5 and 6:

• Configuring Azure ML Pipelines and components for KenAI Automate training, inference, and drift pipelines
• Deploying model using KenAI reusable components
• Performing model testing and MRO dashboard using KenAI Assure
• Ensuring the model is easy to decipher and explain the outcome to the business users leveraging KenAI Govern Setting up KenAI Monitor pipelines with selected model and datasets for Data Drift, Model Drift and Service health dashboards

The Outcome: 2x faster time to market for implementing AI/ML use cases Observability across models with both business and operational monitoring  Adherence to Responsible AI by ensuring models are ethical and easy to decipher Simplified end-to-end Machine Learning Operations (MLOps) with CI/CD/CT/CM