Models, software, and production systems
Machine Learning Engineer
Combine software foundations with the modeling workflow needed to build reliable machine-learning and AI applications.
By the end
Implement model-backed services, reason about their performance, and make evaluation and serving decisions explicit.
Start here
Start with Python for Data Work Python for Data Work
Turn Python records into trustworthy, analysis-ready tables with NumPy and pandas.
Build the software foundation
Learn runtime analysis and the structures most often used in data and serving code.
- Data Science & MLPython for Data WorkTurn Python records into trustworthy, analysis-ready tables with NumPy and pandas.
- AlgorithmsBig-O NotationCount how much work code performs, then describe how that work grows with the input.
- AlgorithmsArraysLearn how indexed storage works, what each array operation costs, and how one safe pass grows into in-place and matrix algorithms.
- AlgorithmsStringsTreat text as an indexed sequence first, then add immutable output, word boundaries, symmetry, runs, direct search, and grammar one idea at a time.
- AlgorithmsHash Tables and SetsProgress from remembering whether a key exists to storing counts, complements, canonical signatures, and one-to-one mappings.
- AlgorithmsRecursionA function that trusts a smaller call to itself to solve a smaller version of the same problem — with a base case as the trapdoor that stops the falling.
- AlgorithmsBinary SearchBinary search isn't about sorted arrays — it's about any monotonic true/false predicate, and the array is just the most common place one shows up.
Understand data and modeling
Retrieve training data, reason about uncertainty, train models, and engineer features safely.
- SQL & DataFiltering RowsAsk a table for exactly the rows and columns you mean, and nothing else.
- SQL & DataGrouping and AggregationCollapse many rows into one answer per group, and know exactly which rows each number counted.
- SQL & DataJoining TablesPut rows from two tables side by side, keep the ones that should survive, and find the ones with no partner.
- Data Science & MLProbability and Statistical ReasoningBuild probability from cases, then use it to reason about repeated observations and evidence.
- Data Science & MLClassical Machine LearningTake one prediction problem through data cleaning, modeling, deployment, and monitoring.
- Data Science & MLFeature EngineeringTurn raw data into useful ML features, with reliable splits, transformations, and model comparisons.
Design model-backed systems
Study graph dependencies and advanced ML systems before building a serving-oriented project.
- AlgorithmsTrees and Binary Search TreesTrees represent hierarchical relationships. Binary search trees add an ordering rule that makes search, insertion, and deletion follow one root-to-leaf path.
- AlgorithmsHeapsA heap keeps only one promise — instant access to the current best — and gives up full ordering to make that promise cheap to maintain.
- AlgorithmsGraphsMost 'graph problems' aren't handed to you as a graph at all — a grid, a word list, or a set of game states IS a graph the moment you decide what a 'node' and an 'edge' mean.
- AlgorithmsTopological SortTurn directed prerequisites into a valid order, or expose the cycle that makes every order impossible.
- AlgorithmsDynamic ProgrammingA repeatable way to build larger answers from smaller answers and reuse work already completed.
- Data Science & MLApplied AI and LLM SystemsDesign grounded language-model workflows that can explain their evidence, abstain, and be evaluated.
- Data Science & ML · OptionalCollaborative Filtering Recommender SystemsPredict tastes and rank recommendations from a sparse ratings matrix using neighborhoods, latent factors, and honest evaluation.