.ipynb) for every live module are linked below and in the reference-notebooks/ folder of this repository.AI & Machine Learning: Advanced Engineering with Cybersecurity
A 12-module, outcome-based journey that builds one running financial AI system — RR Finance — module by module. Every page here is the theory layer: what each algorithm is, why it exists, what came before it, what else you could use instead, and where it shows up in real financial systems. The matching Jupyter notebook for each module is where you run the code.
One system, built module by module
This is not twelve unrelated demos. Every module adds one capability to the same running system, until Module 12 wires everything into one integrated, secured, deployed product.
The 12 modules
Click a live module to read the theory handbook, or use the notebook button to download the runnable .ipynb straight from this repository's reference-notebooks/ folder.
Data Science + Financial Data Analysis
The trusted foundation: EDA, feature engineering, regression, classification, clustering, anomaly detection, and the first security control — data-poisoning defence.
● LiveDeep Neural Networks
Perceptrons, ANNs, backprop, optimizers, CNNs — plus the first inference-time attack: adversarial examples (FGSM/PGD) and adversarial training as a defence.
● LiveNLP + Financial Text AI
Text representations, embeddings, NER, sentiment analysis on financial text, prompt injection, and PII redaction.
● LiveGenerative AI, LLMs & RLHF
Transformers, fine-tuning (LoRA/QLoRA), alignment (SFT → RM → PPO/DPO), and LLM red-teaming.
● LiveRAG, LangChain & AI Agents
Vector retrieval, agentic tool-use, LangGraph, and the agentic-AI threat surface (retrieval poisoning, tool-call hijacking).
● LiveExplainable & Responsible AI
SHAP, LIME, counterfactuals, fairness auditing, and model governance — including the age-fairness question Module 1 flagged.
● LiveFederated & Privacy-Preserving ML
Federated averaging, differential privacy, and Byzantine-robust aggregation defences.
● LiveMultimodal AI
Speech, vision-language models, and document intelligence — building on Module 2's CNN foundations.
● LiveGraph Neural Networks
Fraud-ring detection, knowledge graphs, GraphRAG, and GNN-based intrusion detection.
● LiveMLOps & On-Premises Deployment
FastAPI serving, Docker/Kubernetes, and DevSecOps controls (Trivy, Vault, TLS/mTLS).
● LiveAI Cybersecurity
STRIDE and MITRE ATLAS threat modelling, then real attacks on the deployed model — extraction, evasion, and membership inference — each paired with a measured defence.
● LiveFinal Capstone
Everything from Modules 1–11, applied to real FDIC bank data and wired into one integrated, explainable, secured, deployable product — the CAMEL Sentinel project.
● LiveDownloads
Modules 1–11 each have a matching Jupyter notebook in reference-notebooks/ — run it to build and train the system, and read the handbook page to understand why each piece exists first. Module 12, the capstone, ships as a complete runnable project (CAMEL Sentinel) instead of a notebook.
📓 Notebooks
module1.ipynb · module2.ipynb · module3.ipynb · module4.ipynb
Open in Jupyter, VS Code, or Google Colab. Run top to bottom once before class.
📊 Dataset
rr_finance_module1_dataset.csv
The RR Finance loan-risk table, validated in Module 1 and reused by every module after it.
🐫 Capstone project
CAMEL Sentinel on GitHub · Docker image
The whole course as one deployable app. Run it with a single docker run — see Module 12.
🗂 Repository layout
index.html, module1.html, module2.html — the handbook.reference-notebooks/ — runnable .ipynb files.data/ — datasets.
How to use this handbook
1. Read the theory
Each module page explains what an algorithm is, why it exists, its history, its alternatives, and where it shows up in real financial systems — before you touch code.
2. Build it yourself
Use the hands-on lab guide to type each step into your own blank notebook, one explained piece at a time — the copy-paste-run flow that actually builds muscle memory.
3. Compare against the reference
Download the matching .ipynb to check your work or move faster, then move to the next module — every module saves its artifacts for the next one to load.