Online or in-house trainings within Europe
All workshops in English or German
Full days or half days
Content and schedule adapted to your needs

Build neural networks and transformer architectures from scratch with hands-on PyTorch coding.

This is why you should participate in the workshop (for developers)

• Avoid weeks of scattered tutorials and fragmented online resources to understand how modern AI architectures like transformers and GPT models actually work
• To learn on your own is not for everyone, instead join an interactive session with an expert trainer and a group of participants
• Boost your career by mastering PyTorch and understanding the fundamental building blocks of cutting-edge AI systems like ChatGPT

This is why you should participate in the workshop (for decision makers)

• Thinking about building AI capabilities in-house? Enable your team to understand and implement state-of-the-art neural architectures without months of self-study
• Reduce dependency on black-box solutions by empowering your developers to build, customize, and debug deep learning models from the ground up
• Educate your team and your company with the latest deep learning technologies in order to stay competitive

Schedule

3 full days or 6 half days

Description

In this course, you’ll learn to build neural networks from scratch using PyTorch, starting with fundamentals and progressing to implementing the transformer architectures that power modern AI systems. We’ll begin with PyTorch tensors and automatic differentiation, establishing a solid foundation for understanding how neural networks learn. Rather than treating deep learning as a black box, you’ll implement each component yourself through hands-on coding exercises, from simple perceptrons to multi-layer networks, gaining deep insight into how backpropagation and gradient descent work under the hood.

We will explore specialized neural architectures designed for different types of data and problems. You’ll implement Convolutional Neural Networks (CNNs) for image processing, learning about convolution layers, pooling operations, and feature extraction. For sequential data, we’ll build Recurrent Neural Networks (RNNs), GRU, and LSTM networks, understanding how these architectures handle time-series and text data. Through practical coding, you’ll learn when to apply each architecture type and how to design networks for specific use cases.

The course culminates in implementing the complete transformer architecture from the ground up, the foundation of modern AI systems like ChatGPT. You’ll code encoder-decoder models, implement attention mechanisms from scratch, and understand how self-attention and multi-head attention work together. We’ll cover positional encoding, layer normalization, and feed-forward networks, building a complete transformer step by step. Finally, you’ll implement a simplified GPT-style decoder-only model for text generation, gaining practical understanding of the architecture behind large language models.

Throughout the workshop, we emphasize hands-on PyTorch coding with real exercises in every module. You’ll learn best practices for training and optimizing neural networks, handling computational graphs efficiently, and debugging model implementations. By the end, you’ll have the skills to read AI research papers, implement state-of-the-art models, and understand how modern AI technologies work at a fundamental level, empowering you to build and customize deep learning solutions for your specific needs.

Contact me for a quote

< See all workshops