Ctrl+K

Systems and Toolchains for AI Engineers

Course information

Lectures

Lecture 1: The AI-engineering landscape and the modern toolchain
- L1 demo: an analysis that does not reproduce
Lecture 2: Reproducible environments, version control, and experiment hygiene
- L2 demo: from an empty directory to a tracked run
Lecture 3: Relational databases and SQL for engineering time-series
- L3 demo: a month of sensor data in PostgreSQL
Lecture 4: Columnar storage, Parquet, and DuckDB
- L4 demo: the same data, stored columnar
Lecture 5: Dataframes and batch pipelines
- L5 demo: the same pipeline, two ways
Lecture 6: Streaming concepts and data validation
- L6 demo: validating a live plant stream
Lecture 7: Features for time-series models
- L7 demo: building an ARX regressor matrix
Lecture 8: The machine learning workflow I, time series
- L8 demo: forecasting reactor pressure, and scoring it honestly
Lecture 9: The machine learning workflow II, regression
- L9 demo: regression with scikit-learn
Lecture 10: The machine learning workflow III, classification, experiment tracking and hyperparameter search
- L10 demo: is the plant faulty?
- L10 demo: a hyperparameter search, tracked and registered
Lecture 11: Tensors, autodiff, training loops, and GPUs
- L11 demo: a gradient by hand, a loop by hand, and three ways to break it
- L11: four models that looked fine
Lecture 12: Architectures for engineering data: MLP, CNN, and sequence models
- L12 demo: four architectures for turbofan remaining-useful-life

Assignments

Projects

Search

Ctrl+K

By John Kitchin and Victor Alves