AWS Machine Learning Specialty (AWS) case studies
AWS Machine Learning Specialty · Put your knowledge to work in context.
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 AWS Machine Learning Specialty case studies are 5 cases and about 4,109 words.
1Fraud Scoring: When the Online and Offline Feature Paths Diverge
A mobile payments company scores card transactions in real time. The endpoint must return a fraud score in under 100 ms.
878 words, about 4 min2Churn Model With 0.94 AUC and 0.04 Precision in Production
A subscription business has 1.2M customer-months of data with a 2.1% churn rate and 140 candidate features. The team runs a random 80/20 split with XGBoost and sees validation AUC 0.94 and PR-AUC 0.62.
790 words, about 4 min3Tuning 60 XGBoost Configurations With a Six-Hour Budget
A team must train a tabular fraud model from 200 GB of CSV in S3: 60M rows, 900 features, mostly sparse. They use the SageMaker built-in XGBoost container on a single ml.c5.4xlarge.
809 words, about 4 min4A 250 ms SLO, a Three-Week Traffic Surge, and a Monthly Release Train
A retail bank serves a real-time credit-risk model on a SageMaker endpoint with four ml.m5.xlarge instances behind an Application Auto Scaling target-tracking policy set to 60% CPU utilisation. p99 latency is 180 ms against a 250 ms SLO.
835 words, about 4 min5Weekly 30-Hour Training on Spot Instances That Keep Getting Interrupted
A media company trains a recommendation model for about 30 hours on 8 ml.p3.16xlarge instances and retrains weekly. They adopted Managed Spot Training to control cost.
797 words, about 4 min