AI Engineer (autumn & spring)

Prerequisites for Exchange Students

Minimum 3 successfully completed semesters or an equivalent of 90 ECTS in a Bachelor in Computer Sciences. In this semester prior knowledge and experience is obligatory

Prerequisites for All Students

Statistics:

Being familiar with the following statistical concepts:

• Central tendency: mean, median, mode
• Variability: variance, standard deviation
• Probability distributions: binomial distribution, poisson distribution and the normal distributions
• Correlation (of bi-variate data)
• Simple linear regression
• Doing simple predictions on time series: moving average, exponential smoothing

Online course

Python programming (for data science):

Having basic Python programming knowledge:

• variables
• Python lists & dictionaries
• conditions: if statements
• loops
• Python functions
  • Online courses:
    • https://www.youtube.com/watch?v=rfscVS0vtbw
    • https://www.youtube.com/watch?v=_uQrJ0TkZlc
• Being able to use some data science libraries in Python:
  • Pandas
  • Numpy
  • Matplotlib
  • Online course

You will get an invite to do an online test so the teachers can help you prepare better in case you have some gaps in your prior knowledge We want you to be successful!

Recommended (not obliged) project management experience (eg: WBS, SCRUM, Waterfall, ...)

All students should bring their own laptop, with the following minimum requirements: 16GB RAM memory, i5 processor with VTX and a 64-bit architecture. You can use your preferred operating system: Windows, Linux or Mac OS X.

Click here for Howest language expectations.

More details can be found on www.howest.be/laptops

The course overview above mentions any specific software necessary for the course unit. Howest applies a very strict policy on the use of legal software.

Prerequisites for MLOps course

You need prior knowledge of Docker. Deep learning has to be followed in the same semester as MLOps (or prior knowledge needs to be documented)

Prerequisites for Non-Exchange Students attending as a 1-year short track Bachelor?

Bachelor degree in Computer Science, Applied Informatics or a closely related field.

Have a look at Creative Technologies and Artificial Intelligence – short track

IoT and machine learning are the driving forces of the fourth industrial revolution that is rapidly transforming the world as we know it today. This will inevitably lead to a (nature) shift in the labor market. Many (repetitive jobs) will be taken over by AI applications.

However, there will be enormous opportunities for IT professionals with knowledge of machine learning who can integrate smart algorithms and system.

Particular emphasis is placed on the conceptual understanding of how certain algorithms work. It is important to be able to choose the right machine learning algorithms, train them, evaluate them correctly and improve their performance through hyperparameter tuning.

Themes

• Supervised learning where you learn from labeled data: Linear (multiple) regression with which you can predict continuous outputs. Examples are predicting stock prices, estimating the age of a person based on a picture of the face, predicting risks, making predictions of sales numbers, etc. Classification allows you to divide data into categories. Face recognition, handwriting recognition, cancer detection, predicting whether someone will click on an advertisement or link are just a few examples. Topics and algorithms that are discussed are logistic regression, Support Vector Machines, Naive Bayes, Random Forest Trees and Ensemble learning.
• Unsupervised learning where you get information from non-labeled data. Clustering techniques where you look for similar data. In this way you can discover patterns, relationships and similarities in complex multi-dimensional data. Dimensionality reduction allows us to transform data to the essence. For example, data can be presented more compactly or the performance of machine learning prediction techniques can be increased.
• Neural networks, inspired by how the brain works, allow us to extract insights from data that until recently were not possible. In the machine learning module we look at its conceptual functioning and we build neural networks for regression and classification. This lays the foundation for the deep learning module that builds on this.

The deep learning module continues where the machine learning module has stopped, namely in the neural networks. Applications include natural language processing and sentiment analysis.

Themes

• Repetition of neural neural networks and introduction to deep learning.
• Convolutional Neural Networks (CNN) mainly used for image recognition.
• Auto encoders and restricted Boltzmann machines: reconstruct lost/ damaged data, can be used to generate music or make suggestions.
• Generative Adversarial Networks (GAN). Used for image generation, predicting which medicine will work with certain symptoms, etc.
• Recommendation systems for generating personalized recommendations.
• Neural networks with memory: Recursive Neural Networks (RNN) and Long Short-term memory networks (LSTM): applications include natural language processing and sentiment analysis.
• Reinforcement learning: algorithm learns through interaction with the environment.

In hands-on sessions you’ll acquire the advanced software skills you must master (e.g. containers, Linux…) as an AI Engineer

Through short, lively lesson units, you will learn basic Dutch grammar along with typical expressions and introductory vocabulary to help you during your stay in Kortrijk and Flanders. This language course is complementary to the Strong Belgium Stories lessons and gives additional insights into Flemish and Belgian society and culture.

This course is optional.

This course deepens your knowledge about storytelling by showing you the most characteristic stories of Belgium. Along with our international journalism students, we dive into several “typical” themes, such as heritage and remembrance, arts, European institutions, sports and, of course, chocolate and beer. We explore cutting-edge museums and the European parliament, visit various Belgian cities, be guided by experts, and reflect on Belgian and homeland culture through lectures, visits and assignments related to both tourism and journalism. Please note that this course costs between 100 and 110 euro extra for travel and tours. All excursions are required.

This course is optional.

End goal

In deep learning and advanced AI you learn how we can train different AI models. In this module, the AI services & Mlops module, we want to take these AI models into production. Developers use these trained models in their applications. In addition, AI engineers want to constantly improve and retrain their models.

In this module we will elaborate on deploying Machine Learning and Deep Learning models in a Cloud environment via a CI / CD pipeline. End users call on Microservices or AI services via a frontend environment. The various applications end up in queues that are processed by a scalable back-end environment powered by powerful GPUs. This environment is based on Docker containers and is managed by a Kubernetes platform. The established CI / CD pipeline will be scripted via Python, Powershell or other CLI tools.

The CI / CD pipeline is the backbone of the project. The intention is that this pipeline will be managed by Data Scientists, AI Engineers, front and back end developers and the DevOps team. The entire environment will work on Amazon AWS, Microsoft Azure or Google Cloud Platform. This technically driven module is a deep dive on technology such as Docker, Kubernetes, Kubeflow, Server Less computing platforms. At the end of the module you will be able to set up and manage a complete data pipeline.

You can take up AI for healthcare if you want to study Deep Learning in this specific context.

End Goal

In this course you will learn how to apply the frequently utilized deep learning algorithms in the context of (bio-)medical) data. How to handle patient data, DNA data and medical images such as x-rays are just a few of the topics that will be covered during this course. Another important aspect that will be explored is interpretable AI, as it is often important in the healthcare sector to know why an algorithm takes a certain decision.

Themes

• Introduction to deep learning and neural networks.
• Analysing medical images with Convolutional Neural Networks (CNN), e.g. lung segmentation from x-ray images, and interpretable AI.
• Auto-encoders: can reconstruct lost or damaged data but it can also be used to detect anomalies.
• Generative Adversarial Networks (GAN). Used for, for example, image generation, predicting which medicine will work with certain symptoms, etc.
• The central dogma of molecular biology and interpreting DNA data with Long Short-term memory networks.
• Using reinforcement learning to determine which treatment course is the best option for a patient.

End goal

You complete a research project for a company. The projects are situated in the context of AI & machine learning. Is this only desk research? Of course not: you are required to prove what you state with a working prototype. Challenging!

The machine learning and deep learning modules already provided you with a solid foundation for the concepts of modern AI and how to implement it in a practical way. In this specializing module Advanced AI, we build on these foundations and focus on specific application domains such as advanced computer vision, advanced NLP (natural language processing), generative neural networks, belief networks and multiple-input multiple-output systems.

In addition to further deepening of data-driven machine learning systems, we also study and implement reinforcement learning and deep reinforcement learning systems in this course. Instead of learning from data, these self-learning systems use trial & error to look for an optimal strategy that will give them a maximum reward. These (deep) reinforcement learning strategies mainly find applications in self-learning robots, optimisation of industrial processes, computer games, self-driving cars and personalised recommendations. In addition, we go deeper into a number of popular optimisation and simulation techniques that can significantly improve the performance of your used learning algorithm.

This module is also hand-on with the focus on being able to implement and integrate the AI systems seen in practice. In that respect, it is intended that you come into contact with various state-of-the-art AI frameworks for the development of both deep neural networks and for the design and simulation of (deep) reinforcement learning systems.

The internship is the ideal way to test the practical knowledge and experience. Because a solid company internship requires a certain period of integration, 15 weeks are provided for this. The objective of the internship is twofold: you will gain insight in the business and work on one or more AI business projects or research assignments.

To ensure the smooth running of the internship, an internship agreement is completed and signed by the three parties involved prior to the internship (the internship company, the student and the study programme). The internship agreement clearly states the expected tasks and results of the intern (=the project sheet). In general, we impose three conditions on the internship company: professional supervision, your own workplace and a well-defined assignment of a sufficiently high level.

During the internship period, the student is assigned two supervisors: the internship mentor who is responsible for the daily supervision within the company and the internship supervisor who supervises the student from a distance (based on the internship report).