22 Most Common AWS Combos for Your Cloud AI Deployment

AWS certification exams, especially the Machine Learning Specialty, often focus on recognizing common combinations of services in real-world scenarios. They love asking about how tools integrate, so spotting patterns in these combinations is key to success.

Let’s distill the mock exam patterns into common service combinations likely to appear on the AWS MLS exam, with typical scenarios:

1. Data Ingestion and Preparation Combos

Combination	Pattern in Exam Questions
S3 + Glue + Athena	"How would you query and transform data stored in S3 without moving it to another database?"
Kinesis Data Streams + Lambda + S3	"What services would you use to process real-time sensor data and store it in S3 for analysis?"
Glue + Redshift + QuickSight	"How do you clean data, store it in a data warehouse, and visualize it in a dashboard?"
Glue Data Brew + S3	"Which service allows a visual interface for cleaning and preparing data stored in S3 for ML workflows?"

2. Machine Learning Training and Deployment Combos

Combination	Pattern in Exam Questions
SageMaker + S3	"Where do you store training data for SageMaker jobs?"
SageMaker + Glue	"How would you preprocess large-scale data for SageMaker training?"
SageMaker + Spot Instances	"How can you reduce the cost of training large models in SageMaker?"
SageMaker + Kinesis Data Streams	"What service can feed real-time streaming data into a SageMaker endpoint for predictions?"
SageMaker + SageMaker Neo	"How do you optimize trained models for deployment on edge devices like IoT?"

3. Real-Time Prediction and Streaming Combos

Combination	Pattern in Exam Questions
Kinesis Data Streams + SageMaker Endpoints	"How do you provide real-time predictions for a high-velocity data stream?"
Kinesis Firehose + Lambda + SageMaker	"How would you process incoming events and pass them to a deployed ML model for inference?"
IoT Greengrass + SageMaker Neo	"Which services would you use to deploy and run ML models on IoT edge devices?"

4. Analytics and Visualization Combos

Combination	Pattern in Exam Questions
S3 + Athena + QuickSight	"How can you query and visualize data stored in S3 without moving it to another storage layer?"
Redshift + QuickSight	"Which services would you use to create a business intelligence dashboard on a large dataset?"
Glue + Redshift + SageMaker	"How do you clean data, load it into a data warehouse, and use it to train an ML model?"

5. NLP and Document Processing Combos

Combination	Pattern in Exam Questions
Textract + Comprehend	"How would you extract text from scanned documents and analyze sentiment or entities in the text?"
Textract + Augmented AI	"What service can add human review to document text extraction tasks?"
Translate + SageMaker	"How do you train an ML model for multilingual text classification?"

6. Cost Optimization Combos

Combination	Pattern in Exam Questions
SageMaker + Spot Instances	"How do you save costs on SageMaker training jobs for large datasets?"
Serverless Inference + SageMaker	"Which inference option reduces costs for low-traffic ML applications?"
Multi-Model Endpoints	"How can you host multiple models on a single endpoint to reduce hosting costs?"

How to Recognize These Patterns in the Exam

Look for the core task: Is it training, inference, data cleaning, or visualization?
Identify data flow: Trace the input (e.g., real-time, batch, database) and match services accordingly.
Spot cost or scalability concerns: If the scenario mentions saving costs or scaling, think of services like Spot Instances, Serverless, or Kinesis.