AI news, product launches and company developments.
OpenAI & AnthropicAmazon QuickSight now supports Amazon S3 Tables using Apache Iceberg format as a direct data source. This allows users to query and visualize data stored in S3 without intermediate layers. The feature enables near real-time analytics and simplifies data architecture for AI-ready insights.
Amazon SageMaker AI now offers capacity-aware instance pools for inference endpoints. Users can set a prioritized list of instance types, and SageMaker automatically selects available instances during capacity constraints. This feature works for multiple endpoint types and removes the need for manual intervention.
AWS explains reinforcement learning with LLM-as-a-judge using Amazon Nova models. This method, called RLAIF, improves model fine-tuning by leveraging large language models for evaluation. It enhances training efficiency and model performance.
OpenAI recently changed its partnership with Microsoft, signaling a shift in its business strategy. This move opens opportunities for AWS to strengthen its position in the AI cloud market. The change could impact how AI services are delivered and who leads the market.
AWS has introduced a comprehensive framework to help migrate and upgrade large language models (LLMs) in generative AI production. The solution includes tools and best practices for smooth transitions between different LLMs. This framework aims to optimize prompt conversion and improve model agility in production environments.
Sun Finance implemented an AI-powered identity verification system using Amazon Bedrock, Textract, and Rekognition. This solution increased extraction accuracy from 79.7% to 90.8%, reduced costs by 91%, and cut processing time from 20 hours to under 5 seconds. Combining OCR with large language models enabled more effective ID extraction and fraud detection.
Amazon Quick's agentic AI assistant enhances data analytics by enabling self-service capabilities. It integrates Amazon S3, SageMaker, AWS Glue, and Athena for efficient data querying and management. This approach simplifies analytics workflows and supports multiple storage formats like Iceberg and Parquet.
Amazon Bedrock AgentCore Gateway now supports secure access to private resources via Resource Gateway, which provisions Elastic Network Interfaces inside an Amazon VPC. The blog explains two implementation modes and demonstrates three practical use cases involving private API endpoints and Kubernetes integration. This enables safer and more flexible...
PwC and AWS have developed an AI-driven annotation tool to extract insights from contracts. This tool helps legal and procurement teams quickly find important clauses in lengthy agreements. It addresses the challenge of scaling contract analysis as volumes increase.
AWS Strands Agents significantly reduce AI token usage by up to 96%. This efficiency is achieved through innovative tool design and optimization techniques. Lower token usage means reduced costs and faster AI interactions for developers and businesses.
Vanguard developed a Virtual Analyst solution by emphasizing eight principles of AI-ready data. They utilized various AWS services to implement this system effectively. This approach led to measurable improvements in their business outcomes.
Amazon Bedrock AgentCore Runtime now supports running custom MCP proxies serverlessly. This enables organizations to implement governance, controls, and observability in line with their security policies. The solution simplifies secure management of AI workloads on the cloud.
