Generative engineering course IA

Description of the training

Trains participants to the advanced skills needed to design, develop and deploy models of generative artificial intelligence, covering techniques such as Generating Neuron Networks (GANs), language models (such as GPT), as well as the application of these technologies in various areas such as content creation, image generation, and data synthesis.

Objectives of training

At the end of this training, participants will be able to:

• Understanding the basics of generative AI: gaining in-depth knowledge of the fundamental principles of generative models, such as GANs (Generative Adversarial Networks) and self-regressive models.
• Mastering advanced tools and frameworks: learning how to use technologies such as TensorFlow, PyTorch and APIs from generative models to develop and deploy high-performance solutions.
• Developing practical applications: designing concrete use cases, such as the generation of images, texts, or audio, to meet industrial or creative needs.
• Optimize and deploy generative d💬IA solutions: integrate models into production environments, taking into account scalability, performance and cost constraints.
• Apply ethics and safety in AI solutions: identify and mitigate risks associated with the use of generative models, such as biases, malicious uses or societal impacts.

Who is this training for?

This training is aimed at:

• Software developers: eager to master the tools and frameworks linked to the IA generative to create innovative applications.
• Data Scientists: wishing to deepen their skills in the field of deep learning and advanced generative models.
• Technical architects: seeking to integrate general AI into business solutions and to design adapted architectures.
• IA students and researchers: wanting to explore the latest innovations in generic models for academic or industrial projects.
• IT Professionals in Conversion: Motivated to use general AI to transform creative processes or automate complex tasks.

Prerequisites

No specific prerequisites are required.

Training programme

Day 1: Introduction to the General AI

• Objective: To understand the theoretical bases and main concepts.
• Content:
  Basic concepts of generative AI.
  Overview of generative models: GANs, autoencoders, diffusion models, Transformers.
  Practical applications and use cases.

Day 2: The fundamentals of generative models

• Objective: To explore the mathematical bases and key algorithms.
• Content:
  Introduction to probabilistic distributions and synthetic data generation.
  Supervised vs. unsupervised learning algorithms.
  Case study: generation of tabular data.

Day 3: Exploration of GANs (General Adversarial Networks)

• Objective: To master the operation and implementation of GANs.
• Content:
  GAN architecture: generator and discriminator.
  Problems and optimization of GANs (instability, collapse mode).
  Practical workshop: implementing a GAN for image generation.

Day 4: Various self-regressive models and autoencoders

• Objective: To understand and apply these models for content generation.
• Content:
  Autoencoders: structure and applications (compression, generation).
  Self-regressive models (GPT, LSTM): generation of sequential text.
  Practical workshop: generating creative text with GPT.

Day 5: Deepening Transformers and Deep Learning

• Objective: To explore the basics of Transformers and their role in generative AI.
• Content:
  Transformers principle and attention mechanism.
  Exploration of large models (GPT-3, BERT).
  Workshop: Build a generic model based on Transformer.

Day 6: IA General for Computer Vision

• Purpose: Apply generative models to images.
• Content:
  Introduction to diffusion models (Stable Diffusion, DALL-E).
  Image generation and manipulation.
  Workshop: Generate stylized or realistic images.

Day 7: IA General for audio and video

• Purpose: Explore models for multimedia generation.
• Content:
  Models for the generation of music, audio, and video.
  Introduction to WaveNet and DeepFake.
  Workshop: Voice synthesis with AI tools.

Day 8: Deployment and optimisation of general AI solutions

• Objective: Integrate generative solutions into a real environment.
• Content:
  Optimization of models for production.
  Deployment to the cloud (Azure, AWS, GCP).
  Workshop: deploy a generic model via a REST API.

Day 9: Ethics and safety of general AI

• Objective: Identify risks and adopt responsible practices.
• Content:
  Ethical issues: bias, information manipulation, copyright.
  Security and regulation of generative models.
  Case study: critical analysis of a generic AI project.

Day 10: Completion project

• Objective: Apply the skills acquired in a complete project.
• Content:
  Definition of the project (domain choice: text, image, video).
  Group or individual implementation.
  Presentation and feedback on the solutions developed.

Training assets

• Complete and Progressive Program: A well-defined structure ranging from fundamental concepts to advanced applications for in-depth understanding.
• Practical and Contextual Approach: Many hands-on workshops allow participants to manipulate the tools and models of generative AI in concrete contexts.
• Expertise in Point Tools: Use of the latest and most relevant frameworks and platforms for general AI (Transformers, GANs, Cloud Platforms).
• Development of a Real Project: An entire day devoted to a project of completion of the course, promoting the integration of the acquired into a practical and professional scenario.
• Ethical and Security Dimension: In-depth reflection on ethical issues, biases, and technology regulation to ensure responsible use.
• Adapted to Market Issues: Training designed to meet current business needs for innovative and efficient AI solutions.
• Support and support: Guidance by experts and provision of resources to ensure a sustainable increase in skills.

Pedagogical methods and tools used

• Live demonstrations with data science services.
• Practical workshops and real case studies in various sectors (industry, trade, health).
• Feedback: Sharing best practices and common mistakes in business.
• Simulations and tools: Using simulators for interactive workshops.

Evaluation

• End of training QCM to test the understanding of the concepts addressed.
• Practical case studies or group discussions to apply the knowledge gained.
• Ongoing evaluation during practical sessions.
• Implementation: Complete project from the end of the modules to consolidate the achievements.

Normative References

• Well-Architected Cloud Framework.
• GDPR (General Data Protection Regulation).
• ISO 27001, SOC 2 (Service Organization Control).
• NIST Cybersecurity Framework.