AWS Machine Learning Specialty (AWS) study guide
AWS Machine Learning Specialty · Work through your exam, one topic at a time.
The AWS Certified Machine Learning – Specialty (MLS-C01) is a specialty-level credential aimed at people who can design, build, and operate machine learning workloads on AWS rather than just describe them. It sits above the associate-level cloud certifications and assumes you are already comfortable with core AWS building blocks - compute, storage, networking, and identity - and can reason about statistics and model behavior without hand-holding. The exam is organised around four domains: data engineering (getting data in and out of AWS services and preparing it for training), exploratory data analysis (cleaning, transforming, and visualising data, and choosing useful features), modeling (selecting algorithms, training, tuning, and evaluating models), and machine learning implementation and operations (deploying models, running pipelines, monitoring, and keeping costs and security under control). Expect scenario questions - a business situation with a constraint such as budget, latency, or data volume - rather than pure definitions. The AWS Machine Learning Specialty study guide has 4 sections and about 1,356 words.
01Data Engineering: Ingest, Store, Catalog, Transform
This domain tests whether you can move data from sources into a training-ready store with the right service and cost profile. Use Kinesis Data Streams when you need sub-second latency, custom consumers, multiple readers, or replay (24-hour default retention, extendable to 365 days).
371 words, about 2 min02Exploratory Data Analysis: Finding Problems Before Modeling
EDA questions ask you to diagnose a dataset and choose the correction that will not poison the model. Missing values (rate, pattern, whether missingness is informative), outliers (domain-invalid vs merely extreme), duplicates, target leakage, class imbalance, cardinality of categoricals, skew/transform needs, correlation and multicollinearity, and, for time series, trend, seasonality, and stationarity.
328 words, about 2 min03Modeling: Algorithm Choice, Training, and Evaluation
Modeling is the largest domain. Expect questions that give a data shape and business goal and ask for the algorithm or metric.
346 words, about 2 min04ML Implementation and Operations: Deploy, Monitor, Secure
Most operations questions reduce to one decision: which hosting option matches latency, payload, and traffic shape? Real-time endpoints for persistent, low-latency, up-to-6 MB payloads with auto scaling.
311 words, about 2 min