AI for Leaders & Program Managers: Strategy to Deployment

Explore artificial intelligence fundamentals, the AI life cycle, and generative AI, including AI agents. Build a simple traffic sign classifier and gain hands-on experience with AI tools.

Explore the AI project lifecycle from problem scoping and data acquisition to model building, evaluation, deployment, and ongoing maintenance, with practical self-driving car examples.

Design a healthcare AI project using the seven-step AI lifecycle to predict diabetes risk from health data, covering problem scoping, data collection and exploration, modeling, evaluation, deployment, and ongoing maintenance.

Apply seven-step ai project lifecycle to a healthcare use case, predicting diabetes risk from patient data. Build binary classifiers, deploy to hospital EHRs, and monitor performance for fairness.

Learn how data, a model, and computing power fuel a no-code ai model with Google Teachable Machines for traffic sign classification, covering training, testing, epochs, learning rate, and batch size.

Train an image classification model with Google Teachable Machines to recognize five traffic signs, using training and testing data. Evaluate the model with three sample tests to gauge performance.

Explore extending a Teachable Machine model by adding new classes, uploading 100 images per class, retraining, and mastering the end-to-end lifecycle including architecture, hyperparameter tuning, deployment, and monitoring.

Identify core challenges across the ai project lifecycle, from problem scoping and data handling to modeling, evaluation, deployment, and maintenance, with real-time fraud detection as a use case.

Explore the AI revolution from Eliza to AGI and ASI, tracing milestones like transformers and GPT, and examining levels from conversational AI to autonomous AI.

Explore the ai project lifecycle from problem scoping to maintenance, build a simple model with Google Teachable Machines, and examine data quality, bias, generative ai, ai agents, and deployment challenges.

Define the AI project vision and scope, assess feasibility, and set smart goals. Identify stakeholders, roles and responsibilities, and plan risk, budgeting, and resources for agile or waterfall approaches.

Define a north star vision, scope, and feasibility for AI projects. Align business value with model KPIs and outline an MVP to prevent scope creep.

Assess the technical and organizational feasibility of an AI model to flag dental and outpatient fraudulent health claims at UnitedHealth Group, using BigQuery and Vertex AI for a six-month MVP.

Evaluate technical feasibility using structured data and AutoML to build a dental and outpatient claims MVP, align with leadership, and plan for human in the loop validation and batch predictions.

Define smart goals for AI projects and assess success across model quality, system performance, and business impact, using metrics such as accuracy, F1 score, latency, uptime, and customer outcomes.

Define smart success metrics for an AI project predicting which bank customers will miss monthly loan payments, measuring model accuracy, system reliability, and business impact to proactively reach customers.

Define ai project success with model metrics like 85% accuracy, under 10% false positives, and 90% recall by Q2, while controlling false negatives and reducing missed payments and cost savings.

Explore four AI project stakeholder categories—business and product, data science and AI, development and operations, data engineering—and their roles from AI product manager to data architect.

Define ai project roles in business and product, including ai product manager, industry domain expert, and project manager, linking business, model, and system kpis and trade-offs.

Explore key ai project stakeholders and roles, ai researchers, data scientists, ml engineers, responsible ai and ethics leads, and ux designers, and how they develop, deploy, and govern ai solutions.

Map ai project stakeholders to development, operations, and data engineering roles, and outline pipelines, deployment, and governance for scalable, secure ai systems.

Assign responsibilities for ai powered customer support chatbot, aligning the ai vision with business goals. Distribute tasks across data engineering and data science roles to enable instant answers and escalation.

Define the ai project lifecycle by mapping roles from ai product manager to mlops architect, data scientist, researcher, domain expert, software engineer, ux designer, and project manager in sprint-driven governance.

Compare waterfall and agile development in AI projects, highlighting uncertainty, data dependence, and cross-functional collaboration; agile delivers iterative value with continuous feedback, while waterfall remains linear and rigid.

Compare crisp-dm, agile, and hybrid ai project management frameworks, detailing data understanding, data preparation, modeling, evaluation, deployment, and sprint-driven cycles for rapid prototyping.

Compare crisp-dm, agile, and hybrid approaches in a JP Morgan Chase loan default use case, addressing data understanding, data preparation, feature engineering, risk management, automated decisions, and compliance.

Scale agile with Safe to align cross-functional teams and deliver increments of value through an agile release train, boosting faster time to market, transparency, and continuous improvement.

Identify and mitigate data, technical, deployment, and ethical risks in ai projects through governance, stakeholder alignment, monitoring, and bias audits.

Align technology, people, and teams around a clear business vision for AI project success. Build with robust data quality, governance, multidisciplinary collaboration, agile practices, MLOps deployment, and ethical, responsible AI.

Explore budgeting for a generative ai large language model, analyzing training costs, infrastructure needs, on-premise versus cloud, GPUs like A100/H100, researchers and engineers, energy footprint, and ongoing maintenance.

Estimate frontier model budgeting by analyzing compute, staff, GPUs, interconnects, cooling, and ongoing inference costs, using examples like a 100 million build and 700k daily inference spend.

Master the ai project lifecycle, from vision and mvp to smart goals, stakeholder alignment, and agile or hybrid frameworks, with budgeting and risk management for data quality, bias, and compliance.

Define the data strategy and preparation for AI, covering structured versus unstructured data, labeled versus unlabeled data, data access, quality, feature engineering, cleaning, and governance.

Define a practical data strategy by detailing data acquisition, exploration, cleaning, wrangling, and feature engineering to power AI models and ensure governance, privacy, and KPI alignment.

Apply a nine-pillar data strategy to a predictive maintenance case by integrating sensor data and maintenance logs, ensuring data discovery, quality, governance, privacy, secure storage, labeling, integration, and KPI tracking.

Brainstorm a high-level data strategy for training, validation, and deployment of autonomous vehicle ai, outlining key components and leveraging llama model, lm studio, and gpt four as references.

Define and implement a data strategy for autonomous driving by planning data discovery, collection methods (on-road and simulated), sensor fusion, governance, labeling, privacy and GDPR compliance, and real-time performance metrics.

Define data requirements for training, validating, and deploying AI models by distinguishing structured and unstructured data, and labeled versus unlabeled data, including source, volume, velocity, variety, and ground truth.

Compare supervised learning with labeled data, unsupervised learning with pattern discovery and clustering, and reinforcement learning where an agent learns by interacting with an environment to maximize rewards.

Explore internal data sources owned by your organization and external public datasets and APIs, then learn how to combine them to power the ai project lifecycle.

Explore the Hugging Face data set section, inspect data Studio tools, count datasets by category, and use AI query to filter CNBC news.

Explore external data sources by browsing Hugging Face datasets, filtering news data by most downloaded samples, and using JSON filters to isolate CNBC news for a practical opportunity.

Assess data quality by comparing good versus bad data across relevance, representation, accuracy, completeness, balance, consistency, timeliness, and privacy. See clean, balanced data examples and model performance impact.

Practice data quality assessment by manually cleaning the medical data practice opportunity csv, identifying missing values, duplicates, and vague labels, and adding removal rationales.

Identify and remove data quality issues, including missing ages, missing hospital IDs, unrealistic ages, and duplicates. Emphasize data cleaning and validation to ensure precise, accurate, and trusted artificial intelligence outcomes.

Develop data wrangling and feature engineering techniques to transform messy data into clean, usable features. Apply merging, missing-value handling, one-hot encoding, feature selection, creating new features, scaling, and deduplication.

Explore the pandas library for data analysis and manipulation in Python, including data frames, CSV, Excel, and SQL inputs, and automation with ChatGPT for scalable, repeatable data tasks.

Learn data wrangling and analysis in pandas with python by reading, cleaning, merging client demographics and financial details, handling missing values, and performing sorting, filtering, and visualizations.

Learn to read Excel files into pandas using read_excel, create data frames like df_demographics and df_financials, and visualize data with head, tail, shape, and info to handle missing values.

Learn to handle missing values in pandas by using isnull to detect empties, then fill numerics with the mean and categoricals with the mode.

Learn how to merge two pandas data frames on the client ID using a left join with pd.merge, producing a merged df that combines demographics and financials for analysis.

Master data analysis with pandas by describing a merged dataframe, filtering high net worth clients, and sorting results by investment portfolio size.

Visualize data in Python with Matplotlib, Seaborn, and Plotly to create scatter plots, histograms, and bar charts, revealing age, investment portfolio size, and correlation heatmaps.

Explore data labeling workflows for supervised learning and pricing per label, including image, text, audio, and video annotations by human labelers and automation using Amazon SageMaker Ground Truth.

Compare databases, data warehouses, and data lakes for storage and analytics: databases support transactions, data warehouses enable BI dashboards, and data lakes store structured and unstructured data.

Learn how data pipelines ingest, validate, transform, store and orchestrate data for dashboards and ai models, while applying governance pillars for quality, security and compliance.

Align data strategy with business goals to ensure compliance and minimize bias. Distinguish structured vs unstructured data, labeled vs unlabeled data, and cover supervised, unsupervised, and reinforcement learning.

Explore model development and validation, neural networks and deep learning, training workflows, architectures, KPIs, cloud infrastructure, and transfer learning with a focus on large language models and LMS.

Explore how data, algorithms, and computing power shape AI project lifecycles, from artificial neural networks to deep learning and time-series forecasting.

Explore neural networks and deep learning basics: neurons, layers, and weights, with supervised training across epochs, enabling deep architectures to detect features from edges to faces.

Explore building a simple artificial neural network for binary classification using Google TensorFlow Playground, including neurons, weights, epochs, and learning rate to train and visualize decision boundaries.

Build a neural network with Google TensorFlow Playground, using an 80/20 train/test split on spiral datasets, tuning architecture, hyperparameters like learning rate, and evaluating performance.

Explore classical algorithms and advanced neural architectures, from supervised and unsupervised learning to transformers, GANs, and autoencoders, and learn how to choose the right model architecture for your use case.

Visualize artificial neural networks from input through convolution, downsampling, pooling, and dense layers, and see how features map to outputs in tensor space with AlexNet.

Map problems to neural network architectures—feed-forward, CNNs, RNNs, and LSTMs—and justify the strengths for tasks like electricity forecasting, digit recognition, synthetic face generation, anomaly detection, and English to French translation.

Map neural network architectures to problems, from LSTM for time-series electricity use to CNNs for image classification, GANs for synthetic images, autoencoders for anomaly detection, and transformers for language tasks.

Explore classification models with confusion matrices, true positives, true negatives, false positives, and false negatives, and KPIs like accuracy, precision, and recall to evaluate predictions against ground truth.

Assess a bank's fraud-detection model using a confusion matrix to compute accuracy, precision, and recall. Identify the most critical metric for the bank and explain why.

Compute accuracy, precision, and recall from a confusion matrix for classification models, and learn why precision is prioritized to minimize false positives in fraud detection.

Explore regression models and KPIs to predict continuous values like sales and prices. Learn simple and multiple linear regression, least squares, and how scikit-learn enables forecasting future revenues.

Learn to assess a regression model by measuring residuals and applying MAE, MSE, RMSE, and R-squared, using an ice cream cart revenue example.

Evaluate a Walmart sales forecast using MAE, MSE, and RMSE; explain model reliability and pick a metric for a regional manager, plus a plain-English summary for a VP of sales.

Compute absolute and squared errors to derive MAE, MSE, and RMSE for Walmart store sales forecasts, showing MAE 70k and RMSE 71k as indicators of model reliability.

Learn AI training infrastructure from hardware like GPUs and TPUs to Docker containers and cloud platforms such as AWS SageMaker, Vertex AI, and Azure, with Kubernetes orchestration for scalable training.

Explore experimentation, tracking, and version control in AI projects using weights and biases, MLflow, and git, with real-time dashboards, reproducibility, and production deployment on SageMaker.

Learn how transfer learning repurposes a pre-trained neural network as a starting point, freezing feature extraction layers and fine-tuning a dense, fully connected network to save time and resources.

Explore a practical framework to choose the right AI model for your business. Define needs, pick model type and size, select open or closed sourcing, and plan scalable deployment.

Select the right LM sourcing option—in-house, open-source, or proprietary cloud models—by weighing cost, time to market, data provenance and residency, and supported deployment.

Compare open source large language models on performance using the Hugging Face leaderboard and explain why the old leaderboard became obsolete in favor of the newer, harder benchmarks.

Explore open and closed-source model leaderboards, compare math and physics benchmarks, coding tasks, and open lm score alongside latency and memory to guide model selection.

Celebrate completing the course and showcase your newly acquired skills with pride, practicality, and fun. Explore other courses on artificial intelligence, gen ai, and cloud to continue learning.