Advanced Artificial Intelligence Laboratory

Recommended Prerequisites

Solid knowledge of machine learning concepts and models; Programming Knowledge; Knowledge of the basic concepts of software engineering.

Teaching Methods

The course will promote the understanding of AI and ML operations through a mixture of theoretical and practical learning. In each class, we introduce the basic concepts of each chapter using practical examples. Weekly assignments will focus on deployment tasks, allowing students to apply the concepts in real-world scenarios. There will be a project requiring students to establish a full CI/CD pipeline for an AI model, bridging development and operations, that should consider all the aspects of AI Ops and ML Ops, enabling students to demonstrate their proficiency in managing the end-to-end lifecycle of AI models, from development to deployment and maintenance.

Learning Outcomes

• Develop skills in the effective implementation of AI models.
• Understand and implement ML Ops and AI Ops frameworks.
• Manage the lifecycle of machine learning models from development to production.
• Address challenges such as model drift, scalability, and real-time data integration.
• Tackle challenges related to security and compliance concerning legal aspects.

Work Placement(s)

No

Syllabus

1) Introduction: AI/ML Pipeline; Examples of successful AI systems

2) Data Collection, Preparation and Versioning: Data sources, APIs and web scraping; Data cleaning and transformation; Data Versioning

3) Model Development and Validation: Choice of algorithms and tools; Training, cross-validation and evaluation metrics; Model Versioning

4) Model Implementation: Implementation strategies (cloud, local); Docker and Kubernetes; Pipelines for testing and CI/CD

5) Integration with Applications: API Development; Real-time data processing

6) Scalability and Management: Performance monitoring; Visualizing with Prometheus, Grafana; Scalability for high availability

7) Skid Detection: Detection and correction; Model re-training; A/B Testing

8) Security and Compliance: Privacy and data security; Regulatory compliance

9) Advanced Topics: Federated learning; AI Ops success stories.

Head Lecturer(s)

João Rodrigues de Campos

Assessment Methods

Assessment
Research work: 20.0%
Project: 30.0%
Exam: 50.0%

Bibliography

1.“Machine Learning Engineering” by Andriy Burkov, 2020
2.“ML Ops: Operationalizing Data Science” by David Sweenor, Steven Hillion, Dan Rope, Dev Kannabiran, Thomas Hill, Michael O’Connell, 2020, O’Reilly
3.“Building Machine Learning Powered Applications” by Emmanuel Ameisen, 2020, O’Reilly
4.“Building Machine Learning Pipelines” by Hannes Hapke, Catherine Nelson, 2020, O’Reilly
5.“Managing Data Science” by Kirill Dubovikov, 2019, Packt Publishing
6.“Serving Machine Learning Models: A Guide to Architecture, Stream Processing Engines, and Frameworks” by Boris Lublinsky, O’Reilly Media, Inc. 2017.