Learning AI
Project Overview: From Neurons to Data Centers
This project page documents my deep dive into Artificial Intelligence. Rather than only learning how to write code, I wanted to understand the entire ecosystem, from the mathematical theory of how a machine learns to the physical infrastructure required to run it at scale.
This write-up serves as both a study guide and a design document for a Medical Imaging AI (CNN) project focused on identifying dental health issues.
File 1: AI (Theory and Infrastructure)
This document acts as the foundational textbook for the project. It focuses on the why and where of AI systems.
- Core Concepts: Breaks down the hierarchy of AI, Machine Learning, and Deep Learning, including how neural networks use weights, biases, and backpropagation to learn.
- The Hardware (HPC): Explains High-Performance Computing and how supercomputers, GPUs, and FPGAs physically process data.
- Networking and Storage: Covers the AI data center backbone, including InfiniBand, RDMA, and NVLink for high-throughput parallel processing.
-
Data Center Design: Includes a theoretical pod design for powering a dental diagnostic AI, with network topology and hardware direction.
- Download Word file: AI (Theory and Infrastructure)
File 2: AI Cheat Sheet (Implementation)
This document is the practical how-to guide. It focuses on the code, math, and tuning needed to build usable models.
- Building the Model: Includes Python/Keras snippets for data pipelines, image preprocessing, and layer design.
- Tuning the Brain: Quick reference for hyperparameters like learning rate, batch size, and epochs.
-
Optimization: Summarizes optimizers (for example Adam and SGD) and activation functions (for example ReLU) used to improve learning efficiency and accuracy.
- Download Word file: AI Cheat Sheet (Implementation)
Disclaimer
I am not an AI expert. This page and the attached files reflect my personal learning experience and journey.