Visual Roadmap — Part 1: Overview¶
Foundational diagrams: the big picture, ML paradigms, deep learning architectures, and the NLP/LLM pipeline.
1. The Big Picture¶
flowchart TD
subgraph Foundations
A[Python & NumPy] --> B[Data Science & Pandas]
B --> C[Mathematics & Statistics]
end
subgraph Core ML
C --> D[Tokenization]
D --> E[Embeddings]
E --> F[Neural Networks & Deep Learning]
end
subgraph Applied AI
F --> G[Vector Databases]
G --> H[RAG Systems]
H --> I[MLOps & Deployment]
end
subgraph Advanced
I --> J[Specializations]
J --> K[Prompt Engineering]
K --> L[LLM Fine-tuning]
L --> M[Multimodal AI]
M --> N[Local LLMs]
N --> O[AI Agents]
end
subgraph Production
O --> P[Model Evaluation]
P --> Q[Inference Optimization]
Q --> R[AI Safety & Red-teaming]
end
subgraph Developer Tools
R --> S[VS Code & GitHub Copilot]
S --> T[MCP Servers]
T --> U[Custom Instructions & Agent Config]
end
2. Machine Learning Paradigms¶
flowchart TD
ML[Machine Learning] --> SUP[Supervised Learning]
ML --> UNSUP[Unsupervised Learning]
ML --> RL[Reinforcement Learning]
ML --> SSL[Self-Supervised Learning]
SUP --> CLS[Classification]
SUP --> REG[Regression]
CLS --> LR[Logistic Regression]
CLS --> SVM[SVM]
CLS --> RF[Random Forest]
CLS --> XGB[XGBoost / LightGBM]
REG --> LINR[Linear Regression]
REG --> LASSO[Lasso / Ridge]
UNSUP --> CLUST[Clustering]
UNSUP --> DR[Dimensionality Reduction]
CLUST --> KM[K-Means]
CLUST --> DBSCAN[DBSCAN]
DR --> PCA[PCA]
DR --> TSNE[t-SNE / UMAP]
RL --> MAB[Multi-Armed Bandits]
RL --> PG[Policy Gradient]
RL --> LLM_RL[RL for LLM Post-Training]
LLM_RL --> RLHF[RLHF]
LLM_RL --> GRPO[GRPO]
SSL --> MASK[Masked Language Modeling]
SSL --> CLIP_SSL[Contrastive Learning - CLIP]
3. Deep Learning Architecture Tree¶
flowchart TD
DL[Deep Learning] --> FNN[Feed-forward Networks]
DL --> CNN[CNNs - Convolutional]
DL --> RNN_GRP[RNNs - Recurrent]
DL --> TF[Transformers]
DL --> GAN[GANs - Generative]
DL --> DIFF[Diffusion Models]
FNN --> MLP[MLP / Autoencoders]
CNN --> IMGCLS[Image Classification]
CNN --> OBJ[Object Detection - YOLO]
CNN --> SEG[Segmentation - SAM]
RNN_GRP --> LSTM[LSTM / GRU]
RNN_GRP --> TS[Time-Series Forecasting]
TF --> ENC[Encoder-only - BERT]
TF --> DEC[Decoder-only - GPT]
TF --> ENCDEC[Encoder-Decoder - T5]
TF --> VIT[Vision Transformer - ViT]
TF --> MOE[Mixture of Experts - Mixtral]
GAN --> STYLEGAN[StyleGAN]
DIFF --> SD[Stable Diffusion / FLUX]
DIFF --> SORA[Video Generation - Sora]
4. NLP & LLM Pipeline¶
flowchart LR
A[Raw Text] --> B[Tokenization]
B --> C[Embeddings]
C --> D[Transformer Layers]
D --> E{Task?}
E -->|Generation| F[Decoder - GPT / Llama]
E -->|Classification| G[Encoder - BERT]
E -->|Translation| H[Encoder-Decoder - T5]
E -->|Reasoning| I[Reasoning Model - o3 / R1]
F --> J[Prompt Engineering]
J --> K[Fine-tuning - LoRA / QLoRA]
K --> L[Evaluation - MMLU / HumanEval]
L --> M[Deployment - vLLM / TGI]