CertifyDeckCargando contenido…
The official AWS Certified Machine Learning Engineer - Associate syllabus, organized by domain. Check off each point as you review it.
Inicia sesión para guardar tu progreso entre sesiones.
Collect and store data
SageMaker AI trains on data in Amazon S3, Amazon EFS or Amazon FSx for Lustre: File mode downloads the whole dataset, FastFile mode streams it on demand with file access, and EFS or FSx are mounted through a VPC. To extract data from operational databases without affecting them, DynamoDB export to S3 consumes no read capacity and RDS snapshot export writes Parquet to S3 in the background.
Setting up training jobs to access datasets ↗DynamoDB data export to Amazon S3: how it works ↗
The S3 storage class is chosen by access pattern and resilience: Intelligent-Tiering for unknown access patterns, IA classes for rarely used data (One Zone-IA only if it can be re-created), Express One Zone for the lowest latency and Glacier classes for archives. For regulatory immutability, S3 Object Lock in compliance mode prevents anyone, including the root user, from deleting or overwriting versions during retention.
Understanding and managing Amazon S3 storage classes ↗Locking objects with Object Lock ↗
Ingestion capacity problems often come from poorly distributed keys: in Kinesis Data Streams and in DynamoDB, including global secondary indexes, a low-cardinality partition key creates hot partitions that adding capacity does not fix. In Amazon S3, request rates scale per prefix and gradually, so spread the load across prefixes and retry 503 Slow Down errors.
Choose the right mode to stream in ↗Best practices design patterns: optimizing Amazon S3 performance ↗
Amazon Data Firehose delivers streaming data to Amazon S3 and other destinations without writing applications, buffering it by size or time, and can read directly from Kinesis Data Streams or an Amazon MSK topic. For stateful, time-windowed computation use Amazon Managed Service for Apache Flink, and when several applications read the same stream, enhanced fan-out gives each one dedicated read throughput.
What is Amazon Data Firehose? ↗What is Amazon Managed Service for Apache Flink? ↗
Choose the format by how the data is read: Parquet and ORC are columnar, compress by column and let engines skip blocks, so they reduce the data that Athena and other engines read; convert with CTAS, AWS Glue or Amazon Data Firehose format conversion, which accepts only JSON input. SageMaker AI built-in algorithms require specific formats, such as CSV with no header and the target in the first column.
Use columnar storage formats ↗Common Data Formats for Training ↗
To merge sources in AWS Glue, the Join transform combines two datasets on keys with several join types, and FindMatches links records that share no identifier or exact field values. In Spark, a broadcast hash join avoids a shuffle when one table fits in memory, and AWS Glue job bookmarks make recurring merges process only new data from S3 or JDBC sources.
Record matching with AWS Lake Formation FindMatches ↗Optimize shuffles ↗
Choose the vector store by query pattern: Amazon S3 Vectors provides low-cost vector storage with no infrastructure for infrequent queries, OpenSearch Serverless vector search collections serve high-volume searches with full-text search, filters and aggregations, and Aurora PostgreSQL with pgvector keeps embeddings next to relational data. All of them can serve as the vector store for Amazon Bedrock Knowledge Bases.
Working with S3 Vectors and vector buckets ↗Prerequisites for using a vector store you created for a knowledge base ↗
Amazon S3 stores the original text, image and audio files, and each type is prepared with its own tool: Amazon Transcribe transcribes audio in batch from S3 and can redact PII, and an augmented manifest links images and labels for Pipe mode training. In Amazon Bedrock Knowledge Bases, Bedrock Data Automation turns images, audio and video into text, and a multimodal embeddings model embeds images directly.
Create a knowledge base for multimodal content ↗Augmented Manifest Files for Training Jobs ↗
Amazon SageMaker Feature Store ingests records through the PutRecord API, suited to streaming sources; through the Spark connector for batch loads, where target_stores can write to the offline store only; or through Feature Processing, which manages compute and schedules pipelines. Each record is identified by its ID and event time: the online store keeps only the latest one and the offline store in S3 keeps the full history.
Perform data transformation, feature engineering, and pre-processing
AWS Glue is a serverless data integration service that runs Apache Spark ETL jobs and uses job bookmarks to process only new data. AWS Glue DataBrew cleans and normalizes data without code through reusable recipes, and EMR Serverless runs Spark without managing clusters. SageMaker Data Wrangler, now integrated into SageMaker Canvas, prepares data with little code and exports the flow to S3, Pipelines or Feature Store.
SageMaker Feature Store organizes features in feature groups with a record identifier and an event time. The online store keeps only the latest record for low-latency lookups and supports TTL; the offline store in Amazon S3 keeps the full history for training and batch inference. Training sets without data leakage are built with point-in-time joins on the offline store.
Feature Store concepts ↗Create a dataset from your feature groups ↗
Amazon Data Firehose can invoke a Lambda function that transforms each batch and marks records as Ok, Dropped or ProcessingFailed, and it can convert JSON to Parquet or ORC using an AWS Glue Data Catalog schema. Lambda keeps state across invocations with tumbling windows on Kinesis. AWS Glue streaming jobs use Spark Structured Streaming, process data in windows and use checkpoints instead of job bookmarks.
Transform source data in Amazon Data Firehose ↗Streaming ETL jobs in AWS Glue ↗
Normalization brings numeric values into a [0,1] range and standardization gives them mean 0 and standard deviation 1, so high-magnitude variables do not dominate the model. Binning turns a numeric value into categorical ranges when its relationship with the target is not linear, and a log transformation reduces skew. SageMaker Data Wrangler and AWS Glue DataBrew provide these transformations without code.
Data preprocessing - Machine Learning Lens ↗Feature engineering - Machine Learning Lens ↗
Amazon Bedrock embeddings models turn text, images or both into vectors; multimodal models place them in a shared semantic space so images can be searched by text. A smaller output dimension can be chosen to reduce storage and latency, but the vector index must use the same dimension. Similarity is computed by the vector database, and large indexing volumes are best handled with batch jobs.
Amazon Titan Text Embeddings models ↗Amazon Titan Multimodal Embeddings G1 model ↗
Tokenization splits text into tokens; in SageMaker Data Wrangler a custom tokenizer sets the regex pattern, minimum token length and lowercasing, and the count vectorizer filters terms that are too rare or too common. For LLM data, HTML tags are cleaned, near duplicates are removed with MinHash, and domain-specific synthetic examples can be generated with Amazon Bedrock and reviewed by people in SageMaker Ground Truth.
Transform Data (Amazon SageMaker Data Wrangler) ↗An introduction to preparing your own dataset for LLM training ↗
Amazon Bedrock Knowledge Bases split documents into chunks before embedding: fixed size with overlap, default, hierarchical (retrieves children and returns the parent), semantic, or no chunking; the strategy cannot be changed after connecting the data source. The default parser only extracts text; tables and charts need Bedrock Data Automation or a foundation model parser. A .metadata.json file next to each document adds metadata for filtering.
How content chunking works for knowledge bases ↗Connect to Amazon S3 for your knowledge base ↗
Amazon Macie discovers sensitive data in S3 buckets and generates findings, but it does not modify the data. The AWS Glue Detect PII transform redacts, partially redacts or applies a SHA-256 hash, and can scan every cell or a sample to locate columns. AWS Glue DataBrew offers substitution, shuffling, deterministic or probabilistic encryption, masking and hashing, and Amazon Comprehend redacts PII in text through asynchronous jobs.
Detect and process sensitive data (AWS Glue) ↗Identifying and handling personally identifiable information (PII) ↗
Supervised fine-tuning in Amazon Bedrock uses labeled JSONL files, with prompt and completion fields or the Converse API message format. Continued pre-training uses unlabeled text, with a single text field per line. Model Distillation needs only prompts, or invocation logs filtered by requestMetadata, and the teacher model generates the responses used to fine-tune the student model.
Prepare data for fine-tuning your models ↗Customize a model with distillation in Amazon Bedrock ↗
Validate data quality and manage bias
AWS Glue Data Quality defines rules in DQDL and evaluates them on Data Catalog tables (with rule recommendations) or inside ETL jobs, where it can isolate failing records before loading. Its analyzers and ML-based anomaly detection learn trends and seasonality. In DataBrew, rulesets are validated by profile jobs, which emit events that EventBridge can filter.
AWS Glue Data Quality ↗Validating data quality in AWS Glue DataBrew ↗
Amazon SageMaker Ground Truth labels data with private or vendor workforces, consolidates annotations from several workers and offers active-learning automated labeling for built-in tasks. Verification jobs rate labels and adjustment jobs correct them. To pre-label large volumes with a foundation model, Amazon Bedrock batch inference processes prompts from S3.
Automate data labeling ↗Process multiple prompts with batch inference ↗
SageMaker Clarify pre-training metrics such as Class Imbalance and DPL detect underrepresentation and label differences between groups without a model. To mitigate bias: split by key or with stratification in Data Wrangler, shuffle the data every epoch with ShuffleConfig, and augment images with color, rotation and crop transformations.
SageMaker Clarify computes bias metrics on tabular data: numeric facets are split with a threshold and categorical ones by value. For text or images, build a table of per-item attributes (for example, the language detected with Amazon Comprehend). JS compares multiclass distributions symmetrically, and CDD conditions disparity on subgroups to rule out Simpson's paradox.
Data Wrangler balances classes with random oversampling, random undersampling or SMOTE, which interpolates synthetic samples from numeric features only and copies non-numeric ones. Without changing the data, XGBoost uses scale_pos_weight and multi-label Image Classification uses use_weighted_loss. To evaluate the rare class, use a stratified split and metrics such as balanced accuracy instead of overall accuracy.
Amazon Bedrock fine-tuning data is JSONL: prompt and completion pairs, or the Converse message format for conversational models. The ApplyGuardrail API evaluates text against a guardrail without invoking a model, so violating pairs can be filtered out. Amazon Comprehend detects toxicity and PII (redaction requires an asynchronous job), and Amazon Rekognition moderates images.
Prepare data for fine-tuning your models ↗Use the ApplyGuardrail API in your application ↗
Amazon SageMaker Data Wrangler and AWS Glue DataBrew impute missing values (mean, median, most frequent or last valid value) and detect outliers by standard deviation, quantiles, modified Z-score or IQR, then clip, delete or clear them. Exact duplicates are removed with recipe steps or transforms; fuzzy duplicates without a common ID, with FindMatches in AWS Glue. Random Cut Forest scores multivariate anomalies without labels.
Transform Data - Amazon SageMaker AI ↗Record matching with AWS Lake Formation FindMatches ↗