machine-learning-engineer-roadmap

Advanced Targets

Kernel Programming (C/C++/CUDA) for optimization
optimization techniques for model performance enhancements, including quantization, distillation, sparsity, streaming, and caching.
Integrate and tailor frameworks like PyTorch, TensorFlow, DeepSpeed, and FSDP to advance super-fast model training and inference.
Advance the deployment infrastructure with MLOps frameworks such as KubeFlow, MosaicML, Anyscale, and Terraform, ensuring robust development and deployment cycles.
AI training job scheduling, orchestration, and management via SLURM and Kubeflow.

ML in the real world is about software engineering and Applied Maths

Fundamentals
- What is ML?
  - Input and Output
  - ML Categories
- ML Essentials
  - Linear Algebra
  - Probability & Statistics
  - Calculus (Gradient Descent & Backpropagation)
  - Feature Engineering
  - Loss Functions
  - Evaluation Metrics
- Model Evaluation and Selection
  - Bias-Variance Trade-Off
  - Model Complexity and Overfitting
  - Regularization
  - Interpretability & Explainability
- Model Training
  - Cross Validation
  - Bootstrapping and Bagging
  - Hyperparameter Tuning
  - Training Times and Learning Curves
Supervised ML
- Naive Bayes
- K-Nearest Neighbors
- Decision Tree
- Linear Regression
- Logistic Regression
- Support Vector Machine
Unsupervised ML
- KMeans Clustering
- Singular Value Decomposition
- Gaussian Mixture Model

Data Support / Data Engineering
- Data Ingestion
- Data Storage
- Data Processing
- Processing Orchestration
Model Development
- Workspaces
- Distributed Model Training
- Model Validation (Offline)
Model Production
- Serving and Hosting
- Deployment and Online Evaluation
  - AB Testing (Frequentist & Bayesian)
  - Multi-Armed Bandit
  - Impact Estimation