Phase 0 - Prep and Mental Model

Build math intuition, coding habits, and a full loop: problem -> data -> model -> evaluation.

checkpointprerequisites

Math Basics
Linear algebra, calculus, and probability. Learn with small NumPy exercises.
Coding and Tools
Python 3, virtual environments, pip/conda, and git basics.
Data Literacy
Cleaning data, train/val/test split, and avoiding data leakage.

Tip: run one tiny sklearn project end-to-end first, then deepen theory.

Phase 1 - Machine Learning Fundamentals

Understand supervised learning, loss, optimization, and generalization.

checkpointclassical_ml

Core Concepts
Overfitting/underfitting, bias-variance, and hyperparameter tuning.
Model Families
Linear/logistic regression, tree models, random forests, boosting basics.
Practice
Build one classification and one regression project with clear metrics.

Learn neural networks, backpropagation, tensors, and training dynamics.

checkpointdeep_learning

Neural Basics
Dense layers, activations, softmax, cross-entropy, and learning rate control.
Framework
Pick PyTorch or Keras and understand tensors, autograd, and DataLoader flow.
CNN and Sequence Intro
Basic CNN blocks and sequence modeling concepts for future Transformer work.

Choose 1-2 tracks like CV/NLP and build demo-ready projects.

checkpointdomains

Computer Vision
Classification/detection/segmentation and transfer learning strategy.
NLP
Tokenization, embeddings, and the move toward Transformer-based methods.
Extensions
Time-series, recommender systems, and RL basics.

Prompting, RAG, tool use, evaluation, and safety.

checkpointllm_era

LLM Basics
Attention intuition, pretraining/fine-tuning/alignment concepts, context limits.
RAG Engineering
Embedding, chunking, retrieval, re-ranking, and grounded generation.
Agent Systems
Planning, tools, memory, and defenses against hallucination/prompt injection.

Turn experiments into reliable systems with MLOps and observability.

checkpointproduction

Engineering
Containerization, CI, logs/monitoring, and API service deployment.
MLOps Basics
Experiment tracking, model versioning, rollback, and drift awareness.
Continuous Learning
Read reports/papers, follow ethics/compliance, and contribute to open source.

// Aim for one runnable portfolio artifact in each phase.