Personal Learning Portfolio - AI Governance and Multi-Cloud Deployment Patterns

Open-source experiments and learning projects exploring secure, compliant, and scalable AI systems across multi-cloud environments. This portfolio documents my journey in understanding enterprise AI deployment, governance, and MLOps practices.

🎯 Learning Objectives

• Understanding AI deployment patterns and best practices
• Exploring regulatory compliance considerations (GDPR, EU AI Act)
• Hands-on experience with multi-cloud architectures
• Building reproducible, well-documented solutions

🔧 Technical Skills Demonstrated

• Enterprise AI Agents: Multi-cloud deployment with SAP S/4HANA, IBM watsonx, Google Vertex AI
• Intelligent Search: Retrieval-Augmented Generation (RAG) with vector databases
• ML Operations: Complete MLOps lifecycle with experiment tracking and model registry
• Quality Assurance: Automated drift detection and model performance monitoring
• Security First: Secure supply chain, container hardening, least-privilege access

Personal learning experiments and open-source contributions

k8s-ai-agent-multicloud-rag-iac - Multi-Cloud RAG Agent
Production RAG-powered AI agent with custom document search and Infrastructure-as-Code deployment across AWS, Azure, GCP, and IBM Cloud. Includes FastAPI, Terraform, and Helm charts.
Technologies: Kubernetes, Terraform, RAG, Bedrock, OpenAI, Helm

k8s-ai-agents-demo - Kubernetes AI Agent Reference
Reference implementation for containerized AI agent services with Qdrant vector store, FastAPI endpoints, and MLflow integration for tracking.
Technologies: Docker, Kubernetes, Qdrant, Watson AI, MLflow

s4hana-ai-agent-openshift - SAP S/4HANA Automation Agent
AI-powered automation for SAP S/4HANA deployment on IBM Cloud PowerVS using IBM watsonx.ai, RAG, and Terraform orchestration on Red Hat OpenShift.
Technologies: IBM Watson, PowerVS, OpenShift, Terraform, FastAPI, RAG

vertex-ai-agentbuilder-demo - Vertex AI Agent Builder
Complete examples for building intelligent agents using Google's Vertex AI Agent Builder and Agent Development Kit (ADK), including deployment patterns.
Technologies: Google Vertex AI, ADK, Agent Engine, Cloud Run

vertex-ai-pipeline-demo - ML Pipeline Orchestration
Production MLOps pipelines on Google Cloud using Vertex AI for model training, evaluation, and deployment automation.
Technologies: Vertex AI Pipelines, Python, MLOps

vertex-ai-cloudrun-demo - Serverless AI Services
Lightweight Python/Flask microservices exposing Vertex AI embeddings (text-embedding-004) and Gemini LLM via HTTP endpoints on Cloud Run.
Technologies: Cloud Run, Gemini, Vertex AI, Docker, Python

k8s-mlflow-rancher-desktop - Local MLOps Stack
Developer-laptop stack with MLflow 3.x on Rancher Desktop (k3s), PostgreSQL backend, and MinIO artifact storage for experiment tracking and model registry.
Technologies: MLflow, Rancher, k3s, PostgreSQL, MinIO, S3

k8s-mlfow-demo - Production MLflow Stack
Production-ready MLflow 3.3.2 deployment on Kubernetes with PostgreSQL tracking store, MinIO artifacts, LLMOps capabilities, and evaluation metrics.
Technologies: Kubernetes, MLflow, Helm, MinIO, PostgreSQL, LLMOps

k8s-evidently-demo - AI Quality Monitoring
Kubernetes deployment for Evidently AI to monitor model drift, data quality, and performance. Supports GDPR/EU AI Act compliance with optional Prometheus/Grafana integration.
Technologies: Kubernetes, Evidently AI, Prometheus, Governance, EU AI Act

k8s-ollama-fastapi-demo - Open-Source LLM Serving
Containerized AI inference with FastAPI and Ollama for local LLM deployment, ready for integration with governance and monitoring tools.
Technologies: FastAPI, Ollama, Kubernetes, Docker, LLM, MLOps

🏛️ Governance by Design
Built-in auditability, traceability, and compliance controls from the ground up

🔒 Security & Isolation
Container hardening, secure supply chains, and least-privilege access patterns

🌐 Multi-Cloud Portability
Deploy anywhere: AWS, Azure, GCP, IBM Cloud, or on-premises infrastructure

📈 Observability
Comprehensive monitoring for model drift, data quality, and performance degradation

🧩 Modular Architecture
Composable Infrastructure-as-Code and Helm charts for flexible deployment

For Learners & Contributors:

1. Browse the Featured Projects above
2. Select a topic that interests you
3. Follow the README documentation for setup and experimentation
4. Fork and adapt for your own learning journey
5. Contributions and feedback welcome!

Portfolio Purpose:

• Each project documents my learning process and technical exploration
• All code is Apache 2.0 licensed for educational and personal use
• Experiments based on industry best practices and public documentation
• Shared for community learning and collaboration

🏦 Financial Services
Compliant AI agents with full audit trails for regulatory requirements

🏭 Manufacturing & SAP
Automated ERP deployments with AI-powered infrastructure orchestration

🏥 Healthcare & Life Sciences
Secure RAG systems for document analysis with GDPR/HIPAA considerations

☁️ Cloud Service Providers
Multi-cloud AI platforms with centralized governance and monitoring

Cloud Platforms: AWS, Azure, GCP, IBM Cloud
Orchestration: Kubernetes, OpenShift, Terraform, Helm
AI/ML: Vertex AI, IBM watsonx, OpenAI, Bedrock, Ollama
MLOps: MLflow, Evidently AI, Prometheus, Grafana
Languages: Python, Docker, FastAPI
Vector Stores: Qdrant
Storage: PostgreSQL, MinIO (S3-compatible)

✅ Hands-On Learning: Real implementations, not just theory
✅ Open Source: Apache 2.0 licensed, transparent and collaborative
✅ Multi-Cloud Exploration: Experience with AWS, Azure, GCP, IBM Cloud
✅ Governance Focus: Understanding compliance and security requirements
✅ Continuous Learning: Ongoing exploration of new technologies
✅ End-to-End: From development to deployment and monitoring

Questions & Discussions: Open an issue in the relevant repository
Contributions Welcome: PRs appreciated - please include documentation
Learning Together: Share your experiences and improvements

All repositories are licensed under Apache License 2.0 for educational and personal use.

Note: This is a personal learning portfolio showcasing open-source experiments and educational projects. All work is based on publicly available documentation, tutorials, and best practices. Not affiliated with any commercial entity.

Personal Portfolio by Marian Ropota | Learning through open-source collaboration