Skip to content

Workshops: MLOps for Edge AI

Dates: 26-28 May 2025 | 16-18 June 2025

This intensive three-day workshop focuses on applying MLOps practices to deploy and maintain machine learning models in industrial and edge AI environments. Unlike traditional MLOps approaches, this workshop uniquely emphasizes the challenges and solutions specific to resource-constrained edge devices and industrial machine learning deployments.

MLOps combines machine learning with DevOps practices, enabling the reliable and efficient deployment, scaling, and maintenance of machine learning models in production. Through hands-on sessions, participants will explore how to implement MLOps workflows tailored for industrial AI. Topics include data preparation, model tracking, monitoring, continuous integration/continuous delivery (CI/CD), and edge deployment.

Participants will gain practical experience with the latest tools, frameworks, and optimization techniques for deploying scalable, reliable, and privacy-preserving AI models. The workshop also integrates edge-specific challenges, such as limited compute power, low-latency requirements, and secure on-device processing, providing an approach to MLOps for real-world applications.

Workshop Goals

By the end of this workshop, participants will:

  • Master the MLOps pipeline, from data versioning to model deployment and monitoring.
  • Build skills in containerization, CI/CD, and distributed model training for scalable deployments.
  • Learn advanced optimization techniques for making AI models resource-efficient and production-ready.

Target Audience

This workshop is designed for:

  • Anyone with a background in Python and basic machine learning concepts who wants to acquire practical MLOps skills.
  • Data scientists and ML engineers interested in deploying machine learning models in industrial or edge environments.
  • Developers and IT professionals aiming to integrate MLOps pipelines into production workflows.

Requirements

Participants must bring their own laptop (minimum 8GB RAM) with administrative access to install software. While a significant portion of the workshop will use hardware provided by Hogeschool VIVES (e.g., edge devices and servers), participants will also need to install and run certain tools locally.

Dates and Location

The workshop will be offered twice, with limited attendance to ensure a focused and interactive experience:

First Edition

  • Monday, 26 May 2025 (9:00 - 16:00): MLOps Fundamentals
  • Tuesday, 27 May 2025 (9:00 - 16:00): Building ML Infrastructure
  • Wednesday, 28 May 2025 (9:00 - 16:00): Edge AI Deployment

Second Edition

  • Monday, 16 June 2025 (9:00 - 16:00): MLOps Fundamentals
  • Tuesday, 17 June 2025 (9:00 - 16:00): Building ML Infrastructure
  • Wednesday, 18 June 2025 (9:00 - 16:00): Edge AI Deployment

Location

  • Hogeschool VIVES, Xaverianenstraat 10, 8200 Brugge
  • Main building, classroom B303 (corridor B, 3rd floor)

Day 1: MLOps Fundamentals

This foundational session introduces the MLOps pipeline, focusing on data and model versioning while addressing techniques to monitor and detect model drift.

Morning Session

Introduction to MLOps:

  • Why MLOps is essential for real-world machine learning.
  • Overview of the MLOps pipeline, from development to deployment.

Data Preparation and Versioning:

  • Data preprocessing techniques using Pandas and Notebooks.
  • Visualization of data for insights and monitoring with Grafana and Plotly.
  • Versioning datasets for reproducibility.

Model Tracking and Versioning:

  • Experiment tracking and model versioning using MLflow.
  • Best practices for managing model lifecycle in production.

Afternoon Session

Monitoring and Drift Detection:

  • Monitoring data quality, model quality, and detecting statistical drift.
  • Introduction to autoencoders for advanced drift detection.
  • Strategies for handling drift when detected.

Day 2: Building ML Infrastructure

This session explores creating scalable, automated machine learning infrastructures for deploying and maintaining robust industrial AI systems.

Morning Session

Containers and Virtual Machines:

  • The roles of containers and hypervisors for creating scalable environments.
  • Comparing public cloud and on-premise deployments for industrial use cases.

Infrastructure Management:

  • Infrastructure-as-Code for automating deployment environments.
  • Basics of container orchestration and scaling in Kubernetes.
  • Practical deployment strategies for production ML models.

Distributed Model Training:

  • Frameworks for scalable model training (Ray, Pytorch Lightning).
  • Efficient use of distributed resources for machine learning.

Afternoon Session

CI/CD Pipelines for ML:

  • Building automated workflows with GitHub Actions.
  • Integrating data pipelining frameworks (Prefect).

Model Serving and Monitoring:

  • Deploying and maintaining models in production environments.
  • Monitoring performance and reliability in real-world use cases.

Tracking performance with Prometheus:

  • Collecting metrics from models and infrastructure.
  • Creating dashboards to visualize performance and identify issues.

Day 3: Edge AI Deployment

This final day focuses on optimizing and deploying machine learning models on edge devices, addressing the unique challenges of edge environments.

Morning Session

Introduction to Edge AI:

  • Benefits and challenges of deploying AI on edge devices.
  • Model optimization techniques: quantization, pruning, layer fusion, and knowledge distillation.
  • Introduction to deployment frameworks for edge inference.

Hands-On Deployment with ONNX Runtime:

  • Optimizing pre-trained AI models for edge hardware.
  • Measuring latency, throughput, and performance.

Afternoon Session

IoT Integration with MQTT and REST APIs:

  • Using MQTT for messaging in Edge AI applications.
  • Implementing REST APIs for model inference and device interactions.

Advanced Optimization and Deployment with TensorRT:

  • Converting models to TensorRT for advanced optimization.
  • Optimization of real-time classification and object detection.

Registration and Fees

Participants can register for one, two, or all three days of the workshop, depending on their interests and objectives. Each day is standalone, yet the sessions complement one another for those attending all three days.

Workshop Fees

Companies in the TETRA MLOps4ECM project can send unlimited employees free of charge.

For other companies, the following fees apply:

  • Per Day: €300 per participant.
  • Group Discount: €200 per day for additional participants from the same organization.

Example: The first participant attending all three days pays €900, while a second participant attending the same days pays €600, for a total of €1,500 for two participants.

How to Register

Please fill out the form at: https://forms.office.com/e/pbEPVpCwWF

Or email [email protected] for further information.

Registration includes:

  • Access to workshops for the selected days.
  • Use of provided hardware during sessions.
  • Sandwiches or similar for lunch.
  • Coffee breaks and refreshments.

Seats are limited to ensure personalized guidance and interaction. We encourage early registration to secure your spot!

Cancellation Policy

Cancellation must be done through email: [email protected]

If cancelling up to 7 days before the start of the course, no fee will be charged.

If cancelling less than 7 days before the start of the course, the full fee is due.