Oracle Machine Learning 26ai

6 Saturdays (9am to 11am) | Intermediate to Advanced | Course Notes

Course Content

Week 1 — Introduction to OML 26ai & OML4SQL

• Overview of Oracle Machine Learning (OML) architecture within Oracle AI Database 26ai
• Understanding in-database machine learning and its advantages over traditional ML frameworks
• Introduction to OML4SQL: building ML models using SQL and PL/SQL APIs
• Setting up OML Notebooks and connecting to Autonomous Database and on-premises environments
• Navigating the OML user interface and exploring built-in ML algorithms
• Data exploration and preparation using OML4SQL queries
• Understanding model lineage: tracking models using the BUILD_SOURCE column in data dictionary views
• Best practices for secure and scalable in-database ML development

Week 2 — OML4Py (Python) & Advanced Data Preparation

• Introduction to OML4Py: architecture, proxy objects, and embedded Python execution
• Connecting OML4Py to the Oracle AI Database and configuring Python environments
• Leveraging the native VECTOR data type for similarity search and feature representation
• Advanced data preparation: handling high-cardinality categorical features using the ODMS_EXPLOSION_MIN_SUPP setting
• Scaling and normalizing large datasets for efficient model training
• Data partitioning strategies for training, testing, and validation
• Handling missing data, outliers, and feature engineering at scale
• Performance benchmarking and optimization of data preparation workflows

Week 3 — Supervised Learning: Classification & Regression with XGBoost

• Building supervised learning models using OML4SQL and OML4Py
• Implementing classification algorithms: Logistic Regression, Decision Trees, and Random Forest
• Deep dive into Extreme Gradient Boosting (XGBoost) for classification and regression tasks
• Model evaluation metrics: Accuracy, Precision, Recall, F1-Score, and RMSE
• Hyperparameter tuning and cross-validation techniques within the database
• Interpreting model results using built-in explainability features
• Model scoring and prediction using SQL and Python interfaces
• Managing model persistence and versioning in the database catalog

Week 4 — Unsupervised Learning, Time Series & Feature Extraction

• Implementing clustering algorithms: K-Means, Hierarchical Clustering, and anomaly detection
• Feature extraction using Non-negative Matrix Factorization (NMF) for dimensionality reduction
• Time series forecasting with Exponential Smoothing Method (ESM) — automatic model type and hyperparameter selection
• Using Explicit Semantic Analysis (ESA) for unstructured text: generating dense projections (doc2vec embeddings)
• Combining ESA outputs with traditional ML models to improve classification accuracy
• Handling seasonal data and forecast evaluation
• Detecting outliers and anomalies in streaming data using in-database algorithms
• Case studies: Customer segmentation and demand forecasting

Week 5 — Advanced Features: AI Vector Search & ONNX Integration

• Deep dive into the VECTOR data type and AI Vector Search for semantic queries
• Generating and storing embeddings using OML4Py and OML4R
• Combining vector searches with traditional relational, text, JSON, and graph queries in a single SQL statement
• Exporting OML models to the ONNX (Open Neural Network Exchange) format
• Integrating ONNX Runtime for high-performance, cross-platform inference
• Implementing in-memory sharing of external initializers for ONNX models — reducing memory usage and improving scalability for concurrent workloads
• Deploying ONNX models for real-time scoring inside the database
• Performance tuning of vector indexes (HNSW) with concurrent DML operations

Week 6 — MLOps, REST Integration & Capstone Project

• Understanding the machine learning lifecycle: training, deployment, monitoring, and retraining
• Model management and lineage tracking using data dictionary views
• Deploying OML models as REST endpoints for application integration
• Integrating OML with Oracle Analytics Cloud and third-party BI tools
• Monitoring model drift and performance degradation over time
• Automating retraining pipelines using Oracle Scheduler and OML automation APIs
• Implementing security and access control for ML models and data
• Capstone Project: Build an end-to-end machine learning pipeline from data preparation to model deployment for a real-world business problem (classification, forecasting, or recommendation system)

Project: Develop and deploy a complete machine learning solution using OML4SQL and OML4Py, including data preparation, model training, ONNX export, and REST API deployment.

Starting Date: 4th July 2026
Course Fee: £2990