Neural networks power many of today’s most impactful AI applications. Yet most practitioners interact with them through high-level frameworks, making it difficult to truly understand what is happening under the hood.

In this session, we will demystify neural networks by building one from scratch using core Python. Instead of relying on deep learning libraries, we will implement the key building blocks ourselves to see how neural networks learn from data and improve over time. We will translate theory into functional code and develop a clear intuition for forward propagation, backpropagation, gradient descent, and loss optimization.

Beyond theory, we will connect these fundamentals to practical AI development. Once you understand how training works, it becomes much easier to adapt and improve modern AI systems in production. For example, we will explore how techniques such as LoRA (Low-Rank Adaptation) allow companies to efficiently fine-tune large language models without retraining the entire network, dramatically reducing cost and compute requirements.

Having a strong mental model of these mechanisms is extremely valuable in practice. It helps you debug training issues, interpret model behavior, and make better decisions when designing or fine-tuning machine learning systems. By the end of the session, you will understand how a neural network learns step by step, how to implement a simple one yourself, and how these principles can be applied to build and adapt agentic AI solutions that help tackle real-world business problems.