How behnambina/oci-data-science-ai-samples works
Project type and purpose
• This repository is not a single-running application but a comprehensive “examples and tutorials” codebase for Oracle Cloud Infrastructure (OCI) Data Science and AI services.
• Its purpose is to demonstrate usage patterns for OCI Data Science (notebooks, jobs, pipelines), OCI AI Services (anomaly detection, vision, language, document understanding), model catalog and deployment, data labeling, and associated CI/CD workflows.
High-level architecture and main components
A. Code Repository (GitHub)
• Source folders:
– ai_services (sub-folders for anomaly_detection, document_understanding, language, vision)
– notebook_examples, conda_environment_notebooks, labs, use_cases
– data_labeling_examples, model_catalog_examples, model_deploy_examples, model_deploy_examples
– jobs (Python, Java, Shell, TS+JS, custom actions), pipelines, distributed_training
• Config and metadata: README.md, LICENSE, GitHub Actions workflows
B. Developer/User Environment
• Local workstation or OCI Data Science notebook environment
• Clone or pull this repo from GitHub
• JupyterLab/VS Code UI for notebooks and code examples
C. OCI Cloud Services (the targets of the examples)
1. OCI Data Science Service
– Notebook Sessions running Jupyter notebooks
– Jobs: managed container jobs (Python, Java, TS+JS, Shell)
– Pipelines: orchestrated ML pipelines with multiple steps
2. OCI AI Services
– Anomaly Detection
– Vision (image/video analysis)
– Language (translation, classification, NER)
– Document Understanding (text, tables, key-value)
3. Model Catalog & Deployment
– OCI Model Catalog: model artifacts, runtime.yaml, score.py
– OCI Function or container‐based model deployment
– Autoscaling alarms (CPU/memory)
4. OCI Data Labeling Service
– Bulk labeling via Python and Java scripts
5. OCI Object Storage
– Store datasets, model artifacts, logs, sample data
Relationships and data flows
• The developer pulls the repo → runs notebooks or scripts locally / in Notebook Sessions
• Notebooks and scripts interact with OCI SDK or REST API → invoke AI Services or Data Science Service
• Data flows: sample data (local / Object Storage) → processed in notebooks/jobs → models registered in Model Catalog → deployed via Functions or containers → inference requests
• CI/CD: GitHub Actions kick off workflows (e.g., make_readme, create_pr_action)
Architectural patterns and principles
• Modular folder-per-feature organization: each service or use-case isolated in its own subfolder
• Sample-driven learning: each component shows minimal end-to-end pipeline
• Reuse of common patterns: score.py + runtime.yaml for model deployment
• Infrastructure as code: jobrunner.py + Dockerfiles illustrate containerization
Essential elements
• Clear labeled components: GitHub Repo, Developer Notebook, OCI Data Science Service, AI Services, Object Storage, Model Catalog, Deployment