Engenharia de Modelos Inteligentes (EMI) / Engineering of Intelligent Models (EIM)

This repository contains the source code, lecture materials, slides, and practical resources for the course Engineering of Intelligent Models (MLOps & LLMOps), offered in the Postgraduate Programme in Applied Machine Learning. The course focuses on the engineering, operationalisation, and lifecycle management of machine learning systems, covering reproducibility, orchestration, deployment, monitoring, and LLM operations.

Course Overview

(S = Synchronous Class, AS = Asynchronous Class)

Module	Class	Type	Topics
M1	1	S1	Introduction to MLOps
			ML Lifecycle & Production Challenges
			CRISP-ML(Q)
	2	AS1	MLOps Ecosystem Overview
			Experiment Tracking (MLflow)
			Tooling Landscape
M2	3	AS2	Data Versioning & Reproducibility
			Project Structuring with DVC
	4	AS3	Experiment Management
			Configuration Management with Hydra
	5	S2	Governance & Traceability
			Integrating DVC ↔ MLflow
M3	6	AS4	Workflow Orchestration
			Apache Airflow (DAGs)
	7	AS5	Pipeline Automation
			Ingestion, Training & Evaluation Pipelines
	8	AS6	Tracking & Observability
			Logs, Metrics & Model Artefacts
M4	9	S3	Model Lifecycle Management
			Model Registry & Versioning
	10	AS7	Model Serving & Monitoring
			API Development with FastAPI
			Containerisation with Docker
			Drift & Monitoring Concepts
M5	11	S4	Introduction to LLMOps
			LLM Lifecycle & Tracing
	12	AS8	Prompt Management & Evaluation
			Reliability & Responsible LLM Deployment

Learning Objectives

By the end of this course, students will be able to:

• Design reproducible and traceable machine learning systems.
• Implement automated ML pipelines with orchestration mechanisms.
• Deploy and monitor machine learning models in production environments.
• Manage model lifecycle transitions and version control.
• Understand the operational challenges associated with Large Language Models (LLMs).
• Apply MLOps principles to real-world ML engineering scenarios.

Repository Structure

Each module has its own directory containing lecture materials, slides, code examples, and laboratory exercises.

├── M1_MLOps_Foundations/
├── M2_Design_Data_Model_Engineering/
├── M3_Model_Orchestration_Automation/
├── M4_Operations_Deployment_Monitoring/
├── M5_LLMOps_Essentials/

Directory Description

• M1_*: Foundations, lifecycle concepts, tracking introduction
• M2_*: Reproducibility, data versioning, configuration management
• M3_*: Pipeline orchestration and automation
• M4_*: Deployment, serving, monitoring
• M5_*: LLMOps concepts and evaluation

Core Tools Used in This Course

The course adopts widely used industry-standard tools:

• MLflow – Experiment tracking and model registry
• DVC – Data and pipeline versioning
• Hydra – Configuration management
• Apache Airflow – Workflow orchestration
• FastAPI / LitServe – Model serving via REST APIs / LLM-specific serving
• Docker – Containerisation for reproducibility

Recommended Python Environment

To ensure reproducibility and compatibility:

• Python 3.13+
• Virtual environment using conda, venv, or uv
• Core libraries:
  • mlflow
  • dvc
  • hydra-core
  • apache-airflow
  • fastapi
  • uvicorn
  • docker
  • pandas, scikit-learn, prophet, transformers (for LLMs)
• Jupyter Notebook or VS Code for development