AI Engineering

Recap of essential tools for data manipulation, analysis, and visualization.

• Advanced Python for Data Science (OOP, Decorators).
• NumPy for numerical computing.
• Pandas for data manipulation and analysis.
• Matplotlib and Seaborn for data visualization.

Mastering traditional ML models and evaluation techniques.

• Supervised Learning (Regression, Classification, SVMs, Decision Trees).
• Unsupervised Learning (Clustering, Dimensionality Reduction).
• Model Evaluation Metrics (Accuracy, Precision, Recall, F1-Score, AUC-ROC).
• Cross-Validation and Hyperparameter Tuning.

Creating and selecting features that improve model performance.

• Handling missing data and categorical variables.
• Feature scaling, transformation, and creation.
• Feature selection techniques (Filter, Wrapper, Embedded).
• Introduction to Feature Stores.

Building neural networks from scratch with TensorFlow and Keras.

• Introduction to Neural Networks and Backpropagation.
• Building models with TensorFlow 2.x and Keras.
• Activation functions, loss functions, and optimizers.
• Regularization techniques (Dropout, L1/L2) to prevent overfitting.

Building models that can "see" and interpret images.

• Convolutional Neural Networks (CNNs) architecture.
• Image classification and object detection (YOLO, R-CNN).
• Transfer Learning with pre-trained models (VGG, ResNet, EfficientNet).
• Image segmentation.

Building models that understand text and language.

• Text preprocessing and vectorization (TF-IDF, Word2Vec).
• Recurrent Neural Networks (RNNs) and LSTMs.
• Attention Mechanism and the Transformer Architecture.
• Building models with Transformers (e.g., BERT, Hugging Face).

Understanding and using the latest in Generative AI.

• Introduction to Generative AI (GANs, VAEs).
• Understanding LLMs (GPT, LLaMA) and prompt engineering.
• Fine-tuning LLMs for specific tasks (PEFT, LoRA).
• Retrieval-Augmented Generation (RAG) and vector databases.

Managing the ML lifecycle with MLOps principles.

• The MLOps lifecycle vs. DevOps.
• Experiment Tracking with MLflow.
• Data and Model Versioning with DVC.
• ML Model Registry for governance.

Packaging and scaling AI applications.

• Creating reproducible environments with Docker.
• Building Dockerfiles for ML applications.
• Kubernetes fundamentals for MLOps.
• Running ML workloads on a K8s cluster.

Automating the end-to-end ML workflow.

• Introduction to workflow orchestration (Airflow, Kubeflow).
• Building end-to-end training pipelines with Kubeflow Pipelines.
• CI/CD for ML: Automating pipelines with GitHub Actions.
• GitOps principles for managing ML infrastructure.

Serving models for real-time and batch predictions.

• Batch vs. Real-time inference patterns.
• Building and containerizing a model API (FastAPI).
• High-performance serving with Seldon Core or KServe.
• Deployment strategies (Canary, Shadow) on Kubernetes.

Ensuring model reliability and fairness in production.

• Monitoring for Data Drift and Concept Drift.
• Tracking model performance and operational metrics.
• Setting up monitoring with Prometheus, Grafana, and Evidently AI.
• Model explainability (SHAP, LIME) and bias detection.

Using the AWS-managed platform for MLOps.

• Amazon SageMaker Studio overview.
• SageMaker for data prep, training, and experiment tracking.
• Building SageMaker Pipelines for orchestration.
• Deploying real-time and batch endpoints.

Using Microsoft's managed platform for MLOps.

• Azure ML Studio overview.
• Managing assets, data, and compute in Azure ML.
• Building and scheduling Azure ML Pipelines.
• Deploying models to Azure Kubernetes Service (AKS).

Building a CV or NLP model and the automated training pipeline.

• Project definition and data sourcing.
• Building and training a custom Deep Learning model.
• Setting up MLflow and DVC for the project.
• Building the Kubeflow pipeline for training and validation.

Deploying the model to production and setting up monitoring.

• Creating a FastAPI service for the trained model.
• Deploying the model service to Kubernetes using KServe.
• Setting up monitoring for drift and performance with Evidently AI.
• Final project presentation and code review.