Category: Case Studies
-
Revolutionizing Hospital Care: AI-Assisted Patient Monitoring Without Compromising Privacy
Discover how Hypros and Google Cloud developed an AI-assisted patient monitoring system that detects hospital emergencies without compromising privacy. This innovative solution uses low-resolution sensors and advanced AI models to alert staff to falls, delirium onset, and other critical situations, enhancing patient safety while respecting dignity.
-
Amazon PackScan: Revolutionizing Real-Time Sort Center Analytics with AWS Services
Discover how Amazon transformed its logistics operations with PackScan, an AWS-powered platform that reduced data latency from 1 hour to under 1 minute. This real-time analytics solution processes 500,000 scan events per second across 80 sort centers, resulting in 25% increased throughput and 12% reduction in labor hours.
-
Using Amazon Neptune for Real-time Anomaly Detection in Gaming Transactions
Discover how Zupee leveraged Amazon Neptune’s graph database to detect real-time anomalies in gaming wallet transactions. Learn how they overcame relational database limitations to build an integrity system that processes over 1 million daily transactions, identifies suspicious patterns, and ensures incentives reach legitimate users.
-
How Flutter UKI Optimized Data Pipelines with Amazon MWAA
Discover how Flutter UKI transformed their data pipelines by migrating from EC2-based Airflow to Amazon MWAA, managing 5,500 DAGs and 60,000 daily runs with improved stability and reduced operational overhead.
-
How Meta Manages Data Understanding at Scale: A Privacy-First Approach
Meta has developed an innovative approach to data understanding at scale through their Privacy Aware Infrastructure, using a five-step process that includes schematization, metadata prediction, annotation, asset inventory, and continuous maintenance. This system enables Meta to manage millions of data assets while protecting user privacy and driving product innovation.
-
MIT’s SASA Method: Training LLMs to Self-Detoxify Their Language Output
MIT researchers have developed SASA, a method allowing Large Language Models to detoxify their own outputs without retraining. This system creates internal boundaries between toxic/non-toxic subspaces, helping LLMs generate appropriate content while maintaining natural language fluency—similar to how humans develop internal filters for appropriate speech.
-
Instagram’s Journey to Managing 1000+ ML Models
Instagram has successfully scaled its recommendation system to manage over 1000 ML models. This article explores how they built a robust infrastructure through a model registry, streamlined launching process, and innovative stability metrics to maintain high-quality personalized experiences for billions of users.
-
Building an Operational Data Store with AWS Purpose-Built Databases for Finance Applications
Discover how Amazon Finance Automation built a high-performance operational data store using AWS purpose-built databases. Learn how they combined DynamoDB, OpenSearch, and Neptune to handle hundreds of millions of daily financial transactions with millisecond latency while maintaining flexibility and data accuracy.
-
Enhancing Netflix Recommendations with FM-Intent: Predicting User Session Intent
Netflix has developed FM-Intent, an advanced recommendation model that predicts user intent during viewing sessions. By understanding whether users want to discover new content, continue watching shows, or explore specific genres, FM-Intent delivers 7.4% more accurate recommendations than previous systems, creating a more personalized streaming experience.