AWS
30 articles tagged with this topic
Agents need 'lockfiles' too — AI assistants break between updates, not the model
AI assistants breaking between updates isn't a model problem—it's skills, tools, and permissions drifting. AWS and OpenAI are adding version locks.
SageMaker Adds Batch Feature Writes — AWS Tackles Enterprise AI Engineering Debt
AWS adds batch-write (25 records/call) and listing APIs to SageMaker Feature Store—addressing a common engineering gap that derails enterprise AI proj
AWS Chronos-2 Wins Decathlon Deal for 400M Users — Foundation Models Hit Retail
Decathlon swaps custom forecasting models for AWS Chronos-2. Foundation models leave the chat and enter retail supply chains.
After Salesforce Cuts GPU Costs to 1/8, It Hits a New Wall
Salesforce cut GPU costs 8x with new AWS tools, then hit enterprise availability walls. AI's bottleneck is shifting from "expensive compute" to "runs
AWS Bundles AI Creative Workflow End-to-End — Cloud Vendors Eye Orchestration
AWS teams with fal on a creative Agent pipeline: Quick orchestrates, fal offers 1000+ models, MCP standardizes. Cloud vendors now sell orchestration,
AWS brings OpenAI to India — 'Data stays local' is the new AI selling point
AWS launches OpenAI's GPT-5.6 series in India with cross-region inference that keeps all data within Indian borders.
AWS and NVIDIA Cut AI Inference Costs 75% in Healthcare Deployment
Training models is just the start. AWS and NVIDIA cut Heidi Health's ASR GPUs from 16 to 4, saving 75% compute with sub-second latency.
Deepgram Surfaces the Bill — Self-Hosted AI Voice's Last Black Box Opens
Deepgram now reports billing and GPU status from on-prem voice AI deployments to AWS CloudWatch. A real unlock for banks, hospitals, and government.
AWS Unveils Universal Health Check for AI Agents — Any Framework Can Score
AWS launches AgentCore Evaluations, decoupling agent scoring from any specific framework via OpenTelemetry—directly tackling enterprises' pain of inco
AWS rebuilds SageMaker SDK — 'bring your own model' becomes cloud standard
AWS rewrote SageMaker's Python SDK, unifying training and deployment. Cloud competition shifts from 'better models' to 'easier to use.'
AWS Cuts Medical AI Voice Agent to 1¢ Per Call — Agents Start Doing the Math
Natera built a voice appointment agent on AWS Bedrock AgentCore: under 1¢/call, 100% tool-call accuracy in 500 simulations. Agents are doing the math
AWS kills window-switching for AI debugging — Big Tech races Agent's last mile
AI delivers answers in seconds; engineers still switch browsers to verify. AWS's MCP Apps puts dashboards in AI chat — the final puzzle for enterprise
AWS Turns Enterprise Knowledge Management Into an AI Template
AWS releases an RAG-based AI template for capturing veteran expertise — a key standardization signal for manufacturing, healthcare, and energy.
AWS Makes Distributed AI Training One-Click — The Cloud War For Enterprise AI
AWS brings Ray to SageMaker HyperPod, making distributed AI training one-click. Cloud vendors are racing for AI-curious companies with no ML Ops team.
AWS Turns AI Phone Agents Into Ready-Made Templates for Restaurants
AWS publishes a restaurant AI phone-agent template using Amazon Connect and MCP. Real-world replacement still hinges on the scenario.
AWS to Catalog Enterprise AI Agents — Finding Gets Harder Than Building
Amazon launches Agent Registry and ARD spec to unify enterprise AI agents and tools—answering who approves, who calls, how to find them as enterprise
AWS Bedrock RAG Fix: Pre-Filter Docs With Cheap Model—Save Money, Buy Complexity
RAG queries feed Claude Sonnet 5-20 chunks per ask. AWS proposes pre-filtering with cheap Claude Haiku. Saves tokens, adds complexity.
AWS AgentCore Gateway Targets Agent Sprawl — Bottleneck Is Governance, Not Tech
AWS Bedrock AgentCore Gateway centralizes AI Agent tool access — controlling who calls what matters more than how smart agents are.
AWS ADOP: Auditability Over Automation in AI Data Pipelines
AWS's ADOP reference architecture runs AI through data engineering, but keeps models out of production — revealing real enterprise AI adoption concern
AWS Made ML Drag-and-Drop. The Tutorial Quietly Skips Model Evaluation
AWS released a SageMaker Canvas tutorial letting business analysts train fraud-detection models on Snowflake data without code. The bar drops sharply,
AWS Makes ML Drag-and-Drop — Traditional Firms Can Forecast Without Data Teams
AWS links SageMaker Canvas and Snowflake so business users can build prediction models with no code — part of a hyperscaler race to democratize ML.
GPT-5.6 Lands on Bedrock, Cross-Region Inference Enters Enterprise Cloud Race
GPT-5.6 cross-region inference now spans 25+ AWS regions. The model race is shifting from capability demos to delivery and platform choice.
AWS Uses LLMs to Guard FHIR APIs: From Fixed Rules to Behavioral Analysis
AWS uses Amazon Bedrock to protect FHIR APIs: behavioral analysis trims fixed-rule maintenance, but cost and compliance remain unproven.
AWS Search Filters Make Enterprise AI Agents Safe for Finance and Healthcare
AWS adds domain and time filters to Bedrock AgentCore web search. Behind the detail lies enterprise AI's real deployment blocker: data credibility.
Fanatics Splits Support Into a Fleet of AI Agents — Single Models Fall Short
Fanatics deployed multi-agent AI on AWS to handle 40+ Super Bowl complaints every two minutes while juggling 50-state compliance.
AWS Lets AI Agents Pay Themselves — Cloud Race to Own Machine-Spending Infra
AWS turns AgentCore Payments GA, letting AI agents pay for APIs and tools via stablecoins. Model firms now bet on machine-spending infrastructure.
Axonius Isolates AI Agents Per Enterprise — The Next Lesson SaaS Must Learn
Axonius sandboxes AI Agents per enterprise customer; AWS Bedrock AgentCore offers three tenancy models. Running Agents ≠ enterprise-grade deployment.
NVIDIA Pushes Agent-Specific Small Model on AWS — Big Models Don't Need Every Step
NVIDIA deploys an Agent-targeted small model on AWS one-click platform — single-GPU, open-source. Signal: Agent workflow shifting to tiered routing.
AI Agents Turn Into Black Boxes — AWS Monitors Them on Google, Microsoft
AWS now monitors AI Agents on Google Cloud, Azure, and on-prem — a signal that agents have moved from pilot to production, where "what happens after l
Enterprise AI's Growing Cost Mystery — AWS Adds Per-User Accounting to Bedrock
AWS Bedrock launches per-user cost attribution via Athena and CUDOS, formally acknowledging enterprise AI spend has grown large enough to demand dedic