Stage 1 — Month 1: AI Automation
The AI automation landscape: what to automate, what never to automate
Building no-code and low-code workflows with n8n, Make and Zapier
Working with APIs, webhooks, authentication and JSON data
Connecting AI models into business workflows: email, CRM, documents, spreadsheets, chat
Error handling, logging, scheduling and keeping automations alive in production
Stage project: a deployed multi-step automation that removes a real recurring task
Tools & Software: n8n, Make, Zapier, Google Workspace APIs, Postman, OpenAI / Claude APIs,
Airtable, Slack

Stage 2 — Month 2: Python Programming for AI
Python from zero: syntax, data types, control flow, functions, error handling
Object-oriented programming, modules, packages and virtual environments
Data handling with NumPy and pandas: cleaning, reshaping, joining, aggregating
Visualisation with Matplotlib and Seaborn; reading data like an engineer
Version control with Git and GitHub, plus notebooks vs. scripts vs. production code
Stage project: an end-to-end data analysis on a real public dataset, published to GitHub
Tools & Software: Python 3, Jupyter / Google Colab, VS Code, NumPy, pandas, Matplotlib, Seaborn,
Git, GitHub

Stage 3 — Month 3: Machine Learning
How machines learn: supervised, unsupervised and the bias–variance trade-off
Regression, classification, decision trees, random forests and gradient boosting
Clustering, dimensionality reduction and anomaly detection
Feature engineering, pipelines, cross-validation and hyperparameter tuning
Honest evaluation: accuracy, precision, recall, F1, ROC-AUC — and why accuracy lies
Model deployment basics: saving, serving and exposing a model as an API
Stage project: a trained, evaluated and deployed predictive model with a live demo
Tools & Software: scikit-learn, XGBoost, pandas, Optuna, MLflow, FastAPI, Streamlit, Hugging Face
Spaces

Stage 4 — Month 4: Deep Learning
Neural networks from first principles: neurons, layers, activations, loss functions
Backpropagation, gradient descent, optimisers and learning-rate strategies
Regularisation, dropout, batch normalisation and fighting overfitting
Convolutional and recurrent architectures; an introduction to transformers and attention
Transfer learning and fine-tuning pre-trained models on your own data
Training at scale: GPUs, mixed precision, checkpointing and experiment tracking
Stage project: a custom deep learning model trained and fine-tuned on a domain dataset
Tools & Software: PyTorch, TensorFlow / Keras, Google Colab GPU, Hugging Face Transformers,
Weights & Biases, TensorBoard

Stage 5 — Month 5: Computer Vision
Image fundamentals: pixels, colour spaces, filtering and classical OpenCV techniques
Image classification and transfer learning with modern vision backbones
Object detection and segmentation with the YOLO family and Segment Anything
Face detection, OCR, pose estimation and multi-object tracking
Video pipelines: real-time inference, streaming and performance optimisation
Vision in the real world: quality inspection, retail analytics, security and medical imaging
Stage project: a real-time vision application running on live camera input
Tools & Software: OpenCV, Ultralytics YOLO, Roboflow, MediaPipe, Segment Anything, Tesseract
OCR, PyTorch Vision

Stage 6 — Month 6: Generative AI
How large language models actually work: tokens, context, embeddings, sampling
Professional prompt engineering, structured outputs and evaluation of generated text
Retrieval-Augmented Generation: chunking, embeddings, vector databases and grounding
Fine-tuning and parameter-efficient adaptation (LoRA / QLoRA) on your own data
Image, audio and video generation with diffusion models; multimodal applications
Cost, latency, safety, hallucination control and responsible deployment
Stage project: a deployed RAG-powered generative application over your own document
set
Tools & Software: OpenAI / Claude / Gemini APIs, LangChain, LlamaIndex, Hugging Face, Pinecone /
Chroma, Ollama, Stable Diffusion, ElevenLabs

Stage 7 — Month 7: Agentic AI
From chatbots to agents: goals, planning, memory, reflection and tool use
Function calling and giving an agent safe, scoped access to real systems
Single-agent and multi-agent architectures; orchestration, hand-offs and supervision
Short-term and long-term memory, state management and context engineering
Building agents with LangGraph, CrewAI and the Model Context Protocol (MCP)
Evaluating, guard-railing, monitoring and costing agents in production
Stage project: an autonomous multi-step agent that completes a real workflow end to end
Tools & Software: LangGraph, CrewAI, AutoGen, MCP, OpenAI Agents SDK, vector stores, n8n,
Docker

Stage 8 — Month 8: Physical AI
Embodied AI: perception, decision and actuation loops in the physical world
Sensors and signals — cameras, LiDAR, IMU, depth — and sensor fusion basics
Robotics foundations with ROS 2: nodes, topics, simulation and control
Reinforcement learning for control, and simulation-to-reality transfer
Edge deployment: model compression, quantisation and inference on constrained hardware
Safety, reliability and testing for systems that move in the real world
Capstone project: a physical or simulated AI system that perceives, decides and acts —
plus your full portfolio review and career session
Tools & Software: ROS 2, Gazebo / Isaac Sim, NVIDIA Jetson, Raspberry Pi, OpenCV, PyTorch,
TensorRT, ONNX Runtime, Arduino