Prabazone Computer Education

The Ultimate AI Building Portal

The high-resolution, day-by-day architect journey[cite: 77].

Phase I: The Blueprint (Months 1-2)

Foundational Engine & Fuel

MONTH 01

Math & Python Foundations

+
Week 1: Linear Algebra – The Data Container [cite: 78]
Day 1-2
Vectors & Scalars: Representing "features" like price and size[cite: 80].
Day 3-4
Vector Operations: Addition & Scalar Multiplication (how AI tweaks data)[cite: 81].
Day 5-6
The Dot Product: Measuring "similarity" between data points[cite: 82].
Day 7
Identity & Transpose: Restructuring data utility[cite: 83].
Milestone: Represent a 3-word sentence as a set of vectors[cite: 84].
Week 2: Matrices & Neural Connections [cite: 85]
Day 8-9
Matrix Multiplication: A "system of filters" for massive datasets[cite: 87].
Day 10-11
Systems of Equations: Solving for model "weights"[cite: 88].
Day 12-13
Inverse Matrices & Determinants[cite: 89].
Day 14
Linear Transformations: Visualizing neural network layers[cite: 90].
Milestone: Manually calculate a 2x2 matrix transformation[cite: 92].
Week 3: Calculus – The Optimizer [cite: 93]
Day 15-16
Derivatives: Learning the slope of error[cite: 95].
Day 17-18
The Chain Rule: How errors at the end are caused by neurons at the start[cite: 97].
Day 19-20
Gradients: Adjusting weight and bias simultaneously[cite: 98].
Day 21
Gradient Descent: Walking the "hill of error" to accuracy[cite: 99].
Milestone: Manual gradient descent calculation on a parabola[cite: 100].
Week 4: Python – The Bridge [cite: 101]
Day 22-25
NumPy & Pandas: Creating Tensors and cleaning "Garbage" data[cite: 103, 104].
Day 26-27
Asyncio: Handling multiple API calls to LLMs[cite: 106].
Day 28-30
Visualization & Capstone: Coding Linear Regression from scratch[cite: 107, 109].
MONTH 02

Classical Machine Learning

+
Week 1: Regression & Metrics [cite: 119]
Day 1-3
Linear & Multiple Regression: Predicting prices based on size/age[cite: 121, 122].
Day 4-5
Logistic Regression: The Sigmoid-based "Yes/No" machine[cite: 123].
Day 6-7
Regularization & Metrics: Precision, Recall, and the F1-Score[cite: 124, 125].
Week 2: Decision Logic [cite: 127]
Day 8-11
KNN & Decision Trees: Entropy and Information Gain logic[cite: 129, 131].
Day 12-14
SVM & Preprocessing: One-Hot Encoding and Scaling[cite: 133, 134].
Milestone: Build a bank "Loan Approval" classifier[cite: 135].
Week 3: Ensemble Learning [cite: 136]
Day 15-16
Random Forests: Using 100 trees to reduce error[cite: 138].
Day 17-21
XGBoost (The Gold Standard) & Hyperparameter Tuning[cite: 139, 142].
Milestone: Top 20% on a Kaggle competition using XGBoost[cite: 143].
Week 4: Unsupervised & Production [cite: 144]
Day 22-25
K-Means Clustering & PCA: "Squashing" data columns[cite: 146, 147].
Day 26-30
Scikit-Learn Pipelines & Capstone: End-to-End Fraud Detection[cite: 148, 151].
Phase II: The Architecture (Months 3-4)

Building the Digital Brain

MONTH 03

Deep Learning & Vision

+
Week 1: The Single Neuron [cite: 170]
Day 1-5
Perceptrons & Activation (ReLU/Sigmoid): The digital "On/Off" switch[cite: 173, 176].
Day 6-7
Hidden Layers: Solving the XOR logic problem[cite: 177].
Week 2: Multi-Layer Networks [cite: 179]
Day 8-14
Forward/Backpropagation & Loss Functions (Softmax/Cross-Entropy)[cite: 181, 184].
Milestone: Build an MLP in PyTorch to classify handwritten digits[cite: 185].
Week 3: Computer Vision (CNNs) [cite: 186]
Day 15-21
Kernels, Padding, Stride, and Max Pooling to "scan" pixels[cite: 189, 190].
Milestone: CNN that distinguishes "Circles" from "Squares"[cite: 192].
Week 4: Real-World Optimization [cite: 193]
Day 22-29
Dropout, Data Augmentation, and Transfer Learning with ResNet[cite: 195, 200].
Day 30
Capstone: Personal Security Camera system[cite: 202].
MONTH 04

Language & Transformers

+
Week 1: The Attention Breakthrough [cite: 221]
Day 1-7
Self-Attention (Q, K, V) and Multi-Head perspectives[cite: 224, 226].
Week 2: Transformer Architecture [cite: 228]
Day 8-14
Positional Encoding & Encoder/Decoder (BERT vs GPT)[cite: 231, 233].
Week 3: Tokenization & Embeddings [cite: 236]
Day 15-21
Sub-word (BPE) strategies and Semantic Vector Spaces[cite: 241, 243].
Week 4: Fine-Tuning Experts [cite: 245]
Day 22-30
PEFT & LoRA: Updating only 1% of weights to specialize models[cite: 249, 253].
Phase III: The Autonomous Frontier (Months 5-6)

Agency & Production

MONTH 05

RAG & Long-Term Memory

+
Week 1: Ingestion & Chunking [cite: 269]
Day 1-7
Load-Chunk-Embed Pipeline and Semantic Splitting[cite: 271, 274].
Week 2: Vector Databases [cite: 277]
Day 8-14
Pinecone/ChromaDB and Metadata Filtering[cite: 281, 284].
Week 3: Advanced Retrieval [cite: 286]
Day 15-21
Hybrid Search, Reranking, and Query Expansion (HyDE)[cite: 288, 291].
Week 4: Evaluation [cite: 293]
Day 22-30
The RAG Triad & RAGAS: Measuring Groundedness[cite: 295, 296, 299].
MONTH 06

AI Agents & MLOps

+
Week 1: Thinking Agents [cite: 316]
Day 1-7
LangGraph: Stateful reasoning loops and checkpoints[cite: 320, 322].
Week 2: AI Hands (Tools) [cite: 324]
Day 8-14
JSON Function Calling and MCP (Model Context Protocol)[cite: 326, 328].
Week 3: Multi-Agent Squads [cite: 330]
Day 15-21
CrewAI Orchestration and "Human-in-the-Loop" gates[cite: 333, 334].
Week 4: MLOps Factory [cite: 336]
Day 22-30
Docker, FastAPI, and Final Startup Capstone[cite: 338, 339, 344].