AWS terminology
A
AI data engineer wiring agents, infra, and unapologetic build logs
Overall domain
[Data Sensitivity Keywords] Keyword → AWS Approach --------------------------------------------- "Sensitive info" → Encryption (S3, KMS) "Personal data" → Anonymization, Masking "HIPAA compliance" → HIPAA-eligible services "PII protection" → Data anonymization "Confidential data" → Encryption at rest/transit [Performance Keywords] Keyword → AWS Solution --------------------------------------------- "Real-time inference" → SageMaker Endpoints "Low latency" → Serverless, Edge deployment "Scalable predictions"→ Auto-scaling inference "High throughput" → Distributed training "Complex computations"→ Managed Spot Training [Model Complexity Keywords] Keyword → AWS Approach --------------------------------------------- "Imbalanced dataset" → Class weighting "Feature engineering" → SageMaker Feature Store "Model interpretability" → SageMaker Clarify "Drift detection" → Model Monitor "Bias mitigation" → Fairness metrics [Cost Optimization Keywords] Keyword → AWS Solution --------------------------------------------- "Reduce training cost"→ Spot Instances "Efficient computing" → SageMaker managed spots "Resource optimization"→ Auto-scaling "Minimize infrastructure"→ Serverless options [Deployment Strategy Keywords] Keyword → Deployment Type --------------------------------------------- "Zero downtime" → Blue/Green deployment "Gradual rollout" → Canary deployment "Risk mitigation" → Shadow deployment "A/B testing" → Traffic shifting "Minimal disruption" → Incremental deployment [Data Processing Keywords] Keyword → AWS Service --------------------------------------------- "Large-scale ETL" → AWS Glue "Data lake" → S3 + Glue "Complex transformations" → SageMaker Processing "Data catalog" → AWS Glue Data Catalog "Cross-source integration" → Step Functions [Model Training Keywords] Keyword → Training Approach --------------------------------------------- "Transfer learning" → Pre-trained models "Domain adaptation" → Fine-tuning "Custom algorithms" → Bring Your Own Model (BYOM) "Hyperparameter tuning"→ SageMaker Hyperparameter Tuning "Few training samples"→ Few-shot learning techniquesSageMaker domains:
[Notebook Instance Keywords]
Keyword → Interpretation
---------------------------------------------
"Collaborative ML dev"→ Jupyter notebook
"Shared environment" → Managed notebook instance
"Reproducible research"→ Persistent workspace
"ML experimentation" → Pre-configured environment
[Training Job Keywords]
Keyword → Interpretation
---------------------------------------------
"Model learning" → SageMaker Training Job
"Distributed training"→ Parallel processing
"Large dataset" → Managed training infrastructure
"Complex algorithms" → Scalable training
[Hyperparameter Tuning Keywords]
Keyword → Interpretation
---------------------------------------------
"Optimal parameters" → Hyperparameter Tuning Job
"Automated optimization"→ HPO Job
"Model performance" → Hyperparameter search
"Efficient tuning" → Bayesian optimization
[Inference & Deployment Keywords]
Keyword → Interpretation
---------------------------------------------
"Real-time predictions"→ SageMaker Endpoint
"Offline processing" → Batch Transform
"Edge device deploy" → Edge Manager
"Hardware optimization"→ Neo
"Inference acceleration"→ Elastic Inference
"Endpoint configuration"→ Inference Recommender
[Data Preparation Keywords]
Keyword → Interpretation
---------------------------------------------
"Data labeling" → Ground Truth
"Visual data prep" → Data Wrangler
"Feature management" → Feature Store
"Reproducible features"→ Feature Store versioning
[Model Monitoring Keywords]
Keyword → Interpretation
---------------------------------------------
"Model drift" → Model Monitor
"Bias detection" → Clarify
"Prediction explanation"→ Clarify
"Automated ML" → Autopilot
"Model quality" → Model Monitor metrics
[Advanced ML Keywords]
Keyword → Interpretation
---------------------------------------------
"Automated model dev" → Autopilot
"Cross-device deploy" → Edge Manager
"Custom hardware" → Neo
"Inference optimization"→ Elastic Inference
3 Real-time Processing, Batch Processing, and Storage/Database domains:
[Real-time Processing Keywords]
Keyword → Service/Interpretation
---------------------------------------------
"Streaming data" → Kinesis Data Streams
"Data ingestion" → Kinesis Data Firehose
"Real-time analytics" → Kinesis Data Analytics
"Distributed messaging"→ Amazon MSK
"Continuous data flow"→ Kinesis Services
[Batch Processing Keywords]
Keyword → Service/Interpretation
---------------------------------------------
"ETL transformations" → AWS Glue
"Big data processing" → Amazon EMR
"SQL querying" → Amazon Athena
"Data workflow" → AWS Data Pipeline
"Distributed computing"→ Hadoop/Spark
[Storage Keywords]
Keyword → Service/Interpretation
---------------------------------------------
"Model storage" → Amazon S3
"Large dataset" → S3 scalable storage
"Shared workspace" → Amazon EFS
"High-performance file"→ Amazon FSx
"Scalable storage" → S3 intelligent tiering
[Database Keywords]
Keyword → Service/Interpretation
---------------------------------------------
"Structured data" → Amazon RDS
"Relational database" → RDS (MySQL, PostgreSQL)
"NoSQL storage" → DynamoDB
"Key-value storage" → DynamoDB
"Caching layer" → ElastiCache
"In-memory processing"→ ElastiCache
[Advanced Processing Indicators]
Keyword → Complex Processing Need
---------------------------------------------
"Unstructured data" → Consider Glue/EMR
"Time-series data" → Kinesis/MSK
"Complex transformations"→ EMR/Glue
"Low-latency access" → ElastiCache
"Scalable data store" → S3/DynamoDB
ML Infrastructure domain
[ML Infrastructure Keywords]
Keyword → Service/Interpretation
---------------------------------------------
"Pre-configured environment" → Deep Learning AMIs
"ML-ready compute" → EC2 ML Instances
"Optimized frameworks"→ Deep Learning Containers
"Custom ML hardware" → AWS Inferentia
"Model optimization" → AWS Neuron SDK
[Hardware Optimization Keywords]
Keyword → Optimization Approach
---------------------------------------------
"High-performance inference" → Inferentia Chip
"Model acceleration" → Neuron SDK
"Low-latency ML" → Custom ML hardware
"Efficient computing" → Optimized containers
[Deployment Readiness Keywords]
Keyword → Infrastructure Capability
---------------------------------------------
"ML framework support"→ Deep Learning Containers
"Scalable ML compute" → EC2 ML Instances
"Quick start ML dev" → Pre-configured AMIs
"Framework flexibility"→ Container-based solutions
[Performance Indicators]
Keyword → Infrastructure Consideration
---------------------------------------------
"Complex model deployment" → Deep Learning Containers
"Hardware-specific optimization" → Neuron SDK
"Consistent ML environment" → Standardized AMIs
"Reproducible ML setup" → Containerized frameworks
5.ML Development Domain
[Model Training Keywords]
Keyword → Interpretation/Concept
---------------------------------------------
"Data splitting" → Training/Validation/Test Sets
"Model generalization"→ Cross-validation
"Prevent overfitting" → Regularization
"Model tuning" → Hyperparameters
"Error measurement" → Loss Functions
[Model Optimization Keywords]
Keyword → Optimization Concept
---------------------------------------------
"Parameter optimization"→ Gradient Descent
"Model learning rate" → Optimization algorithm
"Non-linear transformation"→ Activation Functions
"Model complexity" → Bias/Variance Tradeoff
[Performance Evaluation Keywords]
Keyword → Evaluation Metric
---------------------------------------------
"Classification accuracy"→ Confusion Matrix
"Model prediction quality"→ Precision/Recall/F1 Score
"Binary classification"→ ROC/AUC Curve
"Prediction error" → RMSE/MAE Metrics
[Advanced ML Concept Indicators]
Keyword → Complex ML Consideration
---------------------------------------------
"Model complexity management"→ Bias/Variance Balance
"Performance trade-offs"→ Regularization Techniques
"Prediction reliability"→ Cross-validation Strategies
"Error minimization" → Loss Function Selection
[Detection and Evaluation Keywords]
Keyword → Diagnostic Approach
---------------------------------------------
"Model performance" → Confusion Matrix Analysis
"Classification accuracy"→ Precision/Recall Metrics
"Model discrimination"→ ROC/AUC Curve
"Regression accuracy" → RMSE/MAE Evaluation
MLOps Domain
[Model Lifecycle Management Keywords] Keyword → MLOps Concept/Approach --------------------------------------------- "Model tracking" → Model Versioning "Comparative analysis"→ A/B Testing "Centralized storage" → Model Registry "Workflow automation"→ ML Pipeline "Performance tracking"→ Model Monitoring "Deployment safety" → Deployment Strategies [Deployment Strategy Keywords] Keyword → Deployment Technique --------------------------------------------- "Minimal risk rollout"→ Blue/Green Deployment "Gradual introduction"→ Canary Deployment "Zero downtime" → Incremental Deployment "Performance comparison"→ Shadow Deployment "Traffic management" → Staged Rollout [Advanced MLOps Indicators] Keyword → Complex MLOps Consideration --------------------------------------------- "Model reproducibility"→ Versioning & Tracking "Continuous improvement"→ Model Monitoring "Automated workflows" → ML Pipeline Orchestration "Risk mitigation" → Deployment Strategies [Performance Management Keywords] Keyword → Monitoring Approach --------------------------------------------- "Model drift detection"→ Performance Tracking "Prediction quality" → Model Monitoring "Automated governance"→ Model Registry "Experiment tracking" → Versioning Strategies [Comparative Analysis Keywords] Keyword → Evaluation Technique --------------------------------------------- "Model comparison" → A/B Testing "Performance benchmark"→ Experimental Validation "Iterative improvement"→ Continuous Deployment "Version comparison" → Model Registry AnalysisSecurity and Governance in ML domain
[Access Control Keywords]
Keyword → Security Concept
---------------------------------------------
"Resource permissions"→ IAM Roles
"Least privilege" → IAM Policy
"Identity management" → IAM Authentication
"Restricted access" → IAM Roles
[Encryption Keywords]
Keyword → Security Mechanism
---------------------------------------------
"Data protection" → AWS KMS
"Encryption key mgmt" → Key Management Service
"Sensitive data" → Encryption at rest/transit
"Compliance" → Encryption standards
[Network Security Keywords]
Keyword → Network Protection
---------------------------------------------
"Network isolation" → VPC
"Subnet segmentation" → VPC Subnets
"Traffic control" → Security Groups
"Network protection" → Network ACLs
[Governance Keywords]
Keyword → Governance Approach
---------------------------------------------
"Model tracking" → Model Governance
"Compliance workflow" → Approval Processes
"Audit trail" → Version Control
"Reproducibility" → Governance Frameworks
[Data Management Keywords]
Keyword → Data Tracking
---------------------------------------------
"Data source tracking"→ Data Lineage
"Transformation log" → Metadata Management
"Data provenance" → Audit Trail
"End-to-end tracking" → Comprehensive Lineage
[Compliance Indicators]
Keyword → Compliance Consideration
---------------------------------------------
"Regulatory requirements"→ Governance Frameworks
"Audit readiness" → Comprehensive Tracking
"Risk mitigation" → Security Controls
"Transparent ML" → Detailed Lineage
ML Best Practices:
[Data Quality Keywords]
Keyword → Quality Approach
---------------------------------------------
"Data cleansing" → Preprocessing techniques
"Representative data" → Balanced datasets
"Outlier handling" → Statistical methods
"Data integrity" → Validation processes
[Feature Engineering Keywords]
Keyword → Feature Creation Strategy
---------------------------------------------
"Meaningful predictors"→ Feature selection
"Dimensionality reduction"→ Feature extraction
"Domain knowledge" → Expert-driven features
"Complex relationships"→ Non-linear transformations
[Model Evaluation Keywords]
Keyword → Evaluation Technique
---------------------------------------------
"Comprehensive testing"→ Cross-validation
"Performance metrics" → Holistic assessment
"Model generalization"→ Validation strategies
"Comprehensive analysis"→ Thorough model testing
[Production Monitoring Keywords]
Keyword → Monitoring Approach
---------------------------------------------
"Performance tracking"→ Continuous monitoring
"Drift detection" → Model performance metrics
"Degradation prevention"→ Proactive monitoring
"Real-time insights" → Performance dashboards
[Cost Optimization Keywords]
Keyword → Resource Efficiency
---------------------------------------------
"Resource management"→ Efficient compute allocation
"Cost-effective ML" → Optimization techniques
"Scalable infrastructure"→ Dynamic resource scaling
"Minimal waste" → Efficient model training
[Reproducibility Keywords]
Keyword → Consistency Approach
---------------------------------------------
"Consistent environments"→ Containerization
"Experiment tracking" → Version control
"Repeatable results" → Deterministic processes
"Workflow standardization"→ ML pipeline consistency
[Advanced Best Practice Indicators]
Keyword → Sophisticated ML Consideration
---------------------------------------------
"End-to-end quality" → Comprehensive ML lifecycle
"Continuous improvement"→ Iterative refinement
"Strategic ml development"→ Holistic approach
"Performance optimization"→ Advanced techniques