Ml Engineering Course

Lesson 1 • Python for ML Engineering

Covers Python idioms, numerical computing, and data manipulation libraries used throughout the course. Emphasizes code quality and reproducibility from day one.

Lesson 2 • Development Environment Setup

Configures reproducible local and cloud environments for ML work. Introduces version control and dependency management as professional standards.

Lesson 3 • Mathematics Essentials for ML

Reviews linear algebra, calculus, probability, and statistics at the depth required for model understanding. Connects math concepts directly to algorithm behavior.

Lesson 1 • Data Cleaning and Transformation

Applies systematic techniques to handle missing values, outliers, and inconsistent formats. Builds transformation logic that is testable and reusable.

Lesson 2 • Data Quality and Validation

Implements automated data quality checks and schema validation within pipelines. Introduces data contracts as a reliability mechanism for production systems.

Lesson 3 • Data Sources and Ingestion Patterns

Surveys structured, semi-structured, and unstructured data sources and ingestion strategies. Connects data availability to downstream model quality.

Lesson 4 • Feature Engineering Fundamentals

Constructs informative features from raw data using domain knowledge and statistical methods. Demonstrates how feature quality directly drives model performance.

Lesson 1 • Unsupervised Learning Methods

Applies clustering, dimensionality reduction, and density estimation to unlabeled data. Links unsupervised outputs to downstream supervised tasks.

Lesson 2 • Model Evaluation and Selection

Establishes rigorous evaluation protocols using cross-validation, holdout sets, and metric selection. Prevents data leakage and overfitting through proper experimental design.

Lesson 3 • Support Vector Machines and Kernels

Derives the maximum-margin classifier and introduces kernel methods for nonlinear boundaries. Connects SVM geometry to practical hyperparameter choices.

Lesson 4 • Supervised Learning Fundamentals

Covers linear and logistic regression as foundational supervised models with full mathematical derivation. Establishes bias-variance trade-off as a central design concept.

Lesson 1 • Debugging and Profiling Neural Networks

Provides systematic strategies for diagnosing underfitting, overfitting, and training instability. Introduces profiling tools to identify compute and memory bottlenecks.

Lesson 2 • Neural Network Foundations

Derives forward and backward propagation from calculus, then implements a network from scratch. Grounds framework usage in a solid theoretical understanding.

Lesson 3 • Sequence Models and Transformers

Progresses from RNNs and LSTMs to the Transformer architecture and self-attention. Prepares students for NLP and time-series applications in later chapters.

Lesson 4 • Convolutional Neural Networks

Builds CNNs for image classification and object detection using modern architectures. Demonstrates transfer learning to reduce data and compute requirements.

Lesson 1 • Pipeline Architecture Principles

Introduces DAG-based pipeline design, component isolation, and artifact management. Establishes reproducibility and modularity as non-negotiable engineering standards.

Lesson 2 • Workflow Orchestration Tools

Compares and implements pipelines using leading orchestration frameworks. Covers scheduling, dependency resolution, and failure recovery patterns.

Lesson 3 • Experiment Tracking and Reproducibility

Implements systematic experiment tracking to compare runs, manage hyperparameters, and log artifacts. Links reproducibility practices to regulatory and audit requirements.

Lesson 4 • Feature Stores and Reusability

Designs feature stores to decouple feature computation from model training and serving. Addresses online vs. offline feature consistency as a production challenge.

Lesson 1 • Edge and Embedded Deployment

Deploys optimized models to edge devices and mobile platforms with constrained compute. Covers model conversion pipelines and on-device inference frameworks.

Lesson 2 • Building REST and gRPC APIs

Implements production-grade model APIs using FastAPI and gRPC with proper serialization and error handling. Covers API versioning and backward compatibility.

Lesson 3 • Serving Architecture Patterns

Compares online, batch, and streaming inference architectures and their latency-throughput trade-offs. Guides architecture selection based on business requirements.

Lesson 4 • Model Optimization for Inference

Applies quantization, pruning, and distillation to reduce model size and latency without significant accuracy loss. Targets both CPU and GPU inference scenarios.

Lesson 1 • CI/CD for Machine Learning

Extends software CI/CD practices to ML with model validation gates and automated deployment pipelines. Covers testing strategies specific to ML components.

Lesson 2 • Model Governance and Audit Trails

Establishes model cards, lineage tracking, and approval workflows to satisfy governance requirements. Prepares students for regulated industry deployment contexts.

Lesson 3 • Logging, Tracing, and Observability

Implements structured logging, distributed tracing, and dashboards for full ML system observability. Connects observability data to root-cause analysis workflows.

Lesson 4 • Production Monitoring Fundamentals

Establishes monitoring for model performance, data drift, and system health in production. Distinguishes ML-specific monitoring from standard software observability.

Lesson 1 • Cost Optimization and Resource Management

Applies profiling, spot instances, and mixed-precision training to reduce ML infrastructure costs. Introduces FinOps principles adapted for ML workloads.

Lesson 2 • Real-Time ML with Streaming Data

Integrates ML inference into streaming data pipelines for sub-second decision making. Covers stateful feature computation and exactly-once processing guarantees.

Lesson 3 • Scalable Training Infrastructure

Designs distributed training systems using data parallelism and model parallelism strategies. Addresses communication overhead and fault tolerance in large-scale training.

Lesson 4 • Multi-Model and Ensemble Systems

Architects systems that coordinate multiple models through routing, stacking, and cascading patterns. Balances accuracy gains against latency and operational complexity.