AI Cisco Black Belt Academy

4 types of data analytics to help identify use case for AI

1. Descriptive Analytics: What happened? Identify patterns and trends. 2. Diagnostic Analytics: Why did it happen? Recognize root causes. 3. Predictive Analytics: What is likely to happen? Anticipating future needs. 4. Prescriptive Analytics: What should we do? Optimizing processes and decisions.

MLOps Toolkit

All-in-One ML Platforms: These offers include tools for the whole process, from organizing data to running the final model. Experiment Trackers: They help you compare and return to older versions of your models. Deployment Help: Resources for turning your model into something you can use in the real world. Model Monitoring: Use it to keep an eye on how your model is doing after you release it. AutoML Tools: These tools make building ML models easier, even if you are no expert. Workflow Helpers: For creating easy-to-follow, repeatable steps for your ML projects.

Role of AIOps

Applies artificial intelligence to enhance IT operations, improving system performance and enabling predictive capabilities.

CPMAI Step 3: Model Development and Deployment Planning

Apply iterative development principles for model development and refinement cycles. Choose a deployment model (cloud, hybrid, or on-premise) that considers data security, scalability, and existing infrastructure.

Benefits of AI Consume Model

Automate repetitive and routine tasks Provide unique insights to spur your imagination and drive inspiration Interpret customer data, regardless of its complexity Generate forecasts and accelerate decision-making

AI in business help secure solutions with these benefits:

Automate workloads, identify issues, and predict outcomes. Shield your data and infrastructure against vulnerabilities and unexpected issues. Manage large volumes of data swiftly without leakages or risks. Increase productivity, reduce costs, and enhance the customer experience. Aid smarter decision-making processes and accelerate workaround times.

Examples of Intelligent Automation in AI

Automated Data Preparation Automated AI Model Training Automated AI Model Deployment Automated AI Model Monitoring

Essential Elements of AI Governance

Bias Mitigation Data Privacy and Security Transparency and Explainability Accountability

Bias and Discrimination

Bias and discrimination occur when models are trained with inadequate data that lead to unfair predictions, perpetuating societal inequalities and harming marginalized groups.

Benefits of UltraEthernet

Blazing speed Minimal latency RDMA Efficiency Easy scaling Compatibility Cost-Effective

Cisco AI collaboration capabilities

Bridge Language Gaps Reduce Distractions Improve Visualization Enhance collaboration

Summary of Business Value Framework

Business Value AI Business Investment Business case and ROI

Key points of constraints and enhancing AI performance

CPUs and GPUs NAS fabrics HPC communication protocols High Performance Networks

Cisco Multicloud Connect

Cisco SD-WAN Cloud OnRamp: Seamlessly integrate traffic from all major cloud providers, leveraging the high Cisco networking performance and native security. Multicloud networking: Build network connections across various cloud instances for a seamless networking experience.

Converged Infrastructure Solutions

Cisco Unified Computing System (UCS)-powered FlashStack for AI and FlexPod Datacenter for AI were built jointly with industry partners such as PureStorage and NetApp. Through this collaboration, FlashStack and FlexPod Datacenter combine compute, storage, networking, and virtualization into one system. Customers will appreciate UCS solutions, because they simplify programmatic deployment, intelligent automation, and autonomous management for mid-sized AI workloads.

Hyperconverged Infrastructure

Cisco has partnered with Nutanix and their multicloud platform to deliver the most flexible, resilient, and future-ready UCS-powered hyperconverged solutions. This collaboration supports customers with global operations.

CPMAI Step 1: Business Knowledge

Collaborate with clients to learn the specific problems they are trying to solve, along with their strategic objectives. Assess whether AI is the right solution for the client's challenges. Define success metrics for the AI project.

Key vulnerabilities that commonly affect AI applications

Complexity Sensitive Data Exfiltration Improper Use of Data

Cisco AI Operationalize

Create an operating model that drives AI DevOps by considering how AI integration will fit within daily operations. Or, must the organization create a new operating model based on the new AI capability?

3 Key Ingredients to Success in AI Business

Creativity: AI is fueling a once-in-a-generation boom in innovation across virtually every industry. There are opportunities for you, as a technology advisor to your customers, to co-create new use cases and to help them clarify and accelerate their AI strategy. Feasibility: Every customer will have a unique alignment of ambitions and constraints, and you will be one among many advisors. Be clear about which domains reflect your value proposition, and within those domains, advise your customer diligently on feasibility, risk, and cost. Stakeholder Buy-in: AI is making a very direct and visible impact on user experience across a wide range of job roles and industries. The vendors and partners who will benefit most from this opportunity will be those who engage business stakeholders and user constituencies and who can help their customers align technology investment with business KPIs.

Unique benefits of extended models

Customization Enhanced Performance Scalability Reduced Dependency Regulatory Compliance

Things to consider to operate AI on cloud/on-prem/hybrid

Data Integrity and Generation Infrastructure Scale and Lock-in Network Performance Computation scale Operating costs Power Efficiency

Data Processing Tasks

Data acquisition Data storage Data transformation

Compute Infrastructure Elements

Data and AI pipelines GPUs power AI Compute Architecture Fit Hosting Locale Fit

Observability in AI Monitoring

Data pipelines Model performance

Stages of AI Model

Data preprocessing Model training and fine tuning Model evaluation Model deployment and production AI inferencing service

AI Solutions to the Govt

Deliver enhanced citizen services Create and make data-driven policy decisions Modernize and streamline operations Optimize infrastructure management Explore AI-powered trafficdesign and public safety

TPUs

Developed specifically for AI applications, TPUs are designed for even faster and more efficient tensor operations, a core mathematical function in AI algorithms.

Cisco Multicloud Observe

Digital experience monitoring: Gain complete observability across users, infrastructure, and applications to align with business objectives. Application security: Apply business risk-based security measures for better insights and prioritization. Cost and resource optimization: Optimize costs and resources across multicloud environments without sacrificing performance or reliability.

CPMAI Step 5: Monitoring and Governance

Discuss the importance of continuous monitoring to ensure the AI model performance meets expectations. Implement a model governance framework to manage risks and ensure responsible use of AI.

Data Governance: Compliance

Ensuring that data management practices adhere to regulations and industry standards.

CPMAI Step 2: Data Knowledge

Evaluate the availability, accessibility, and quality of data with the client. Discuss data management issues, including privacy, security, and compliance

Contributions of MLOps

Evolving practices Value Delivery Continuous Improvement

AI Business framework starting point

Experiment POC Prototype Pilot

Benefits of Identifying the Primary Business Constraint

Focused Improvement Strategic Investment Continuous Analysis

Role of MLOps

Focuses on managing the machine-learning model lifecycle, including data preparation, model training, testing, and deployment. It helps ensure that models are scalable and robust.

Cisco AI Personas

For whom are they doing this enablement? Customers? A preferred or critical supplier? A global community? How will the intended audiences benefits correlate to the ROI?

Critical Success Factors to Customers

High-Quality Data Domain Expertise Scalability and Integration Change Management

Cisco AI Investments and Business Case ROI

How do they intend to fill talent gaps, through new talent or talent skill-ups? What changes must they make to infrastructure to support AI integration? When third-party vendors offer new capabilities, are those capabilities in line with their strategy and worth the investment?

Cisco AI Information Security

How will they implement a Secure Development Lifecycle (SDL)? How are they assessing risk? How will they apply zero-trust when collaborating with third-party vendors? Will they follow the National Institute of Standards and Technology (NIST) Risk Management Framework (RMF)?

Cisco AI Data Classification and Privacy

How will they use data and maintain privacy? Different countries have various requirements for privacy. How will they take these varying requirements into consideration?

Reverse Engineering AI Hypothesis

Identify Required Data Determine AI model needs Construct a detailed hypothesis Pilot project implementation Test and refine the hypothesis Continuous monitoring and adjustment

Benefits and advantages of AI in security for customers

Identify and block threats before they cause damage Reduce manual effort and accelerate incident response Optimize security continuously against evolving threats Prevent costly data breaches and reputational harm Minimize security operations workload

Data Governance: Best Practices

Implementing policies and procedures to effectively manage and protect data assets.

Cisco AI Benefits in Observability

Increase revenue and sales conversion Minimize service disruption and downtime Provide unified visibility into complex environments Improve customer experience and retention Detect anomalies and degradations early Accelerate troubleshooting and issue resolution

Classic Observability

Infrastructure Resources Application Performance

AI Solutions to the Manufacturing

Intelligent quality control Proactive machine maintenance Digital twin creation Supply-chain optimization and tracking Optimized production processes

AI governance

It encompasses policies, processes, and best practices that prioritize fairness, transparency, accountability, and alignment with ethical principles. By taking a proactive stance on governance, organizations can mitigate the risks associated with AI and foster an environment where compliance is a natural result.

LangChain

LangChain is an open-source framework for building applications based on large language models (LLMs). Applications: LangChain uses the power of AI LLMs combined with data sources to create powerful apps.

KPIs to watch out for when monitoring your system?

Latency Resource Utilization Availability Accuracy

several ways to use AI to enhance sales effectiveness

Lead Generation Opportunity Scoring AI Driven Sales Forecasting

Observability Tools in MLOps that help

ML Flow: These models track experiments, versions, and deployments, enabling performance comparisons and issue identification. Prometheus and Grafana: These popular tools are for collecting and visualizing metrics from models and infrastructure, offering insights into system health. Shapley Additive explanations: This one measures AI workload components or features and their relationships to rank how important they are to a prediction.

MLOps

Machine-learning operations (MLOps) is the bridge between data scientists and operations. In short, the goal of MLOps is to simplify AI processes and deployment to increase quality, especially at scale.

Data Governance: Ethical Considerations

Making decisions that prioritize privacy, security, and fairness in data management.

AI Solutions to the Healthcare

Medical imaging analysis Enhance diagnosis and treatment Patient management through predictive analytics Improved access to healthcare with remote monitoring tools Drug research and development

Benefits in prioritizing AI Governance

Minimizing Risk Ethical Alignment Fostering Innovation Competitive Advantage

Observability in MLOps

Observability involves collecting, aggregating, and analyzing data about the ML system, including: Model performance: Metrics like accuracy, precision, recall, and F1 score help gauge model effectiveness. Data quality: Monitoring data distributions, drift, and potential biases in production data is crucial. Infrastructure health: Keeping an eye on system resources, errors, and logs helps identify potential bottlenecks or issues with the serving environment.

Operational Inefficiencies

Operational inefficiencies may occur when models are fed with insufficient data that does not allow the ML to make accurate forecasts.

GPUs

Originally designed for graphics processing, GPUs excel at parallel processing tasks, making them well-suited for accelerating AI model training.

Risks behind ill-trained models

Overfitting Bias and Discrimination Operational Inefficiencies

Overfitting

Overfitting occurs in machine learning when a model captures not only the underlying patterns in the training data but also the random fluctuations or noise. This noise leads to a model that performs exceptionally well on its training dataset but poorly on any unseen data, as it fails to generalize from its training experiences.

GPUs are essential because they provide:

Parallel Processing Speed Deep Learning

AI Solutions to the Education

Personalized and tailored learning experience Intelligent tutoring systems Automated grading and assessments Enhanced content development through data-driven insights Greater student and faculty safety

AI Solutions to the Finance

Predictive trading algorithms Fraud detection and prevention Personalized financial advice Investment portfolio optimization Virtual assistants and seamless transaction experiences

AI malpractices

Privacy Violations Bias and Discrimination Manipulation and Deception

Benefits of observability for AI

Proactive issue detection: Continuous monitoring enables proactive identification and resolution of issues, minimizing the impact on end-users. • Faster root cause analysis: Observability aids in pinpointing the source of problems, leading to reduced mean time to resolution (MTTR) and minimized downtime. • Enhanced security: Integration of application performance data with security context facilitates early detection of security risks and blind spots in multicloud environments. • Cross-team collaboration: Observability provides a shared knowledge of AI application health and performance, fostering collaboration across data science, ML engineering, infrastructure, and DevOps teams. • Improved efficiency and productivity: Observability automates tasks, freeing up IT teams to focus on strategic initiatives like model optimization and feature development.

Cisco AI security capabilities solutions

Provide automated and proactive threat detection Conduct behavioral analysis Generate automated responses to threats Generate adaptive security policies Promote an integrated security ecosystem

PyTorch

PyTorch is an open-source ML library from the Facebook AI Research lab. Applications: PyTorch is particularly popular among researchers and developers for its flexibility and dynamic computation graph. Developers use it for applications such as computer vision, natural language processing, and generative modeling.

Key Aspects of Digital Twins in AI

Real-Time Data Integration Predictive Analytics Adaptive Control

Possible AI Benefits to Networking Solutions

Reduce human error and increase efficiency Minimize downtime and improve network performance Adapt to changing network conditions in real time Lower operational costs through automation Increase agility to support dynamic business needs

Cisco Multicloud Secure

Security Service Edge (SSE): Protect user access across the internet, SaaS, and private clouds, ensuring security and consistency. Next-gen multicloud firewall: Implement unified security policies across multicloud environments to safeguard your data and applications.

MLOps Consideration for Partners

Skills and Resources People, Process, Tools Strategic Role Determination Automation and Workflow Integration LAER Framework Application Foundation vs Custom Models

AI Impact on Case Management

Software Recommendations Automated Diagnostic Scans Efficient Case Routing

TensorFlow

TensorFlow is an open-source machine learning library that Google developed. Applications: TensorFlow is one of the most widely used frameworks and is suitable for a wide range of applications, including image and speech recognition, natural language processing, and reinforcement learning.

Embedded Model

This model allows companies to augment current infrastructure with AI functionality.

Extended Model

This model connects to an external service, which reduces infrastructure needs but introduces limitations.

Consume Model

This model does not require extensive infrastructure, making AI more accessible.

Cisco Responsible AI Framework

Transparency Fairness/Impartiality Inclusion Accountability Privacy Security Reliability

CPMAI Step 4: Model Implementation and Operationalization

Use project management best practices for effective implementation and deployment. Plan change management strategies to integrate AI solutions into the client organization.

Cisco AI Use Cases

What are the relevant use cases? Will they be focused on productivity gains and hardening applications and infrastructure? Will they provide BizOps or DevOps improvements?

To adopt AI, you must know the following:

What business processes can it assist? What data can it use? Is it necessary to change any procedures or decision-making processes? Do you need to make personnel changes or adopt a new corporate culture? What other similar questions apply specifically to your situation?

Solutions to AI Security Risks

Zero Trust Segmentation or Microsegmentation Data Loss Prevention

scikit-learn

scikit-learn is a versatile ML library that provides simple and efficient tools for data mining and data analysis. It is widely used for machine-learning applications. Applications: scikit-learn is commonly used for tasks like classification, regression, clustering, and dimensionality reduction.