• AI solution lifecycle overview
• Loading and inspecting real datasets
• Handling missing data, duplicates, outliers
• Data transformation and encoding
• Feature creation and selection
• Using AI copilots for coding efficiency

Lab:
Clean and prepare a business dataset

• Statistical summaries and distributions
• Correlation and feature relationships
• Visual diagnostics
• Identifying data leakage and bias early

Lab:
EDA with insight-driven interpretation

• Supervised learning overview
• Regression vs classification
• Common algorithms:
  • Linear / Logistic Regression
  • Random Forest
  • Gradient Boosting
• Bias-variance trade-off
• When simple models outperform complex ones

Lab:
Train baseline regression and classification models

• Train/test splits and cross-validation
• Regression metrics: RMSE, MAE, R²
• Classification metrics: Accuracy, Precision, Recall, F1, ROC-AUC
• Model comparison and selection
• Avoiding overfitting

Lab:
Compare multiple models and select the best

• Feature scaling and encoding
• Interaction features
• Handling imbalanced datasets
• Hyperparameter tuning (conceptual + basic practice)
• Performance vs explainability trade-offs

Lab:
Improve model performance systematically

• Test sets vs validation sets
• Data drift and model decay
• Grounding AI outputs
• Guardrails for reliability
• Human-in-the-loop design

Case Study:
Model failure and mitigation strategies

• Why deployment matters
• Introduction to Streamlit
• Introduction to Gradio
• Designing user inputs and outputs
• Error handling and validation

Lab:
Deploy a regression model with Streamlit

• Multi-tab applications
• Regression and classification in one app
• Sliders, dropdowns, and user controls
• Input validation and safety checks

Lab:
Build a Gradio app with multiple models

• What is an API?
• Exposing models as REST endpoints
• Connecting AI models to:
  • Dashboards
  • Internal tools
  • Automation pipelines
• When to automate vs when not to

Demo:
AI model → API → dashboard workflow

• Basic monitoring concepts
• Logging and error tracking
• Model performance monitoring
• Security and access control
• Responsible AI in production

• Triggers, actions, and pipelines
• Automating data ingestion
• Scheduling model runs
• Integrating AI into business workflows
• Intro to orchestration tools (conceptual)

Lab:
Design an AI-powered automated workflow

• Translating business problems into AI tasks
• Defining success metrics
• Cost vs performance trade-offs
• Communicating results to stakeholders

Exercise:
AI solution design canvas

Participants will:

• Select a business use case
• Prepare data
• Train and evaluate models
• Deploy using Streamlit or Gradio
• Design an automation flow

Outcome:
End-to-end AI solution

• Demo deployed solutions
• Peer and instructor feedback
• Reliability and risk review
• Roadmap: Builder → AI Engineer / Automation Architect