LoginSign Up
← Back to all sparks
A

AWS Machine Learning

AI-ASSISTANTS
Velocity10.0

Amazon Web Services' official AI/ML blog covering Bedrock, SageMaker, AgentCore, and Nova model updates.

aws.amazon.com/blogs/machine-learning ↗

AWS ML's blog has become an agentic-infrastructure showcase, not a model gallery.

agentic-infrastructurebedrock-agentcoremcpsagemakermodel-availabilitycompliance
◆Current state
The SageMaker and Bedrock content stream now reads almost entirely as agent enablement: AgentCore Runtime for hosting coding agents, Strands Agents for domain reasoning, Amazon Quick orchestrating MCP servers, and Nova Sonic voice evaluation. Model-availability posts like Nemotron 3 Ultra on JumpStart still appear but are outnumbered by infrastructure-for-agents pieces. The throughline is operating agents in production, not just calling models.
◆Where it's heading
AWS is positioning Bedrock AgentCore as the runtime layer for long-running, isolated agent sessions and pushing MCP as the integration substrate across its services. Expect more posts pairing AgentCore with third-party tools like New Relic and Asana, plus compliance-oriented routing such as cross-region inference for the EU.
◆Prediction
The next entries likely deepen AgentCore with managed memory, gateway tooling, or observability, and add more named-model launches on JumpStart.

◆Recent moves

  1. 8h ago

    Scale Robot Reinforcement Learning with NVIDIA Isaac Lab on Amazon SageMaker AI

    A walkthrough for training Unitree H1 humanoid policies on SageMaker HyperPod and Training Jobs. It pushes SageMaker into robotics reinforcement learning, a workload class AWS rarely highlights, fitting the broader move toward running heavier agent and ML training on managed compute.

    View source ↗
  2. 12h ago

    Hands-free first notice of loss: Using Strands Agents and Amazon Bedrock AgentCore Browser Tool for intelligent claims intake

    Demonstrates the AgentCore Browser Tool driving live insurance-portal interaction for claims intake, paired with Strands Agents for reasoning. Another concrete agent-does-the-screen-work pattern in the agentic push.

    View source ↗
  3. 12h ago

    Build an agentic incident triage assistant with Amazon Quick and New Relic

    Amazon Quick orchestrates a New Relic MCP server and Asana to assemble a root-cause brief from one prompt. Shows AWS leaning on MCP as the connective tissue between its agents and external tools.

    View source ↗
  4. 1d ago

    Unlocking AI flexibility in Europe: A guide to cross-region inference for EU data processing and model access

    Cross-Region Inference on Bedrock routes requests across EU regions to balance capacity against data-residency rules. A compliance-shaped capability that matters more as agent workloads scale in regulated markets.

    View source ↗
  5. 1d ago

    It’s safe to close your laptop now: Hosting coding agents on Amazon Bedrock AgentCore

    ⚡ SPARK

    Positions Bedrock AgentCore Runtime as the place to host coding agents (Claude Code, Codex, Kiro, Cursor), each in its own isolated microVM with a persistent workspace. The clearest statement yet that AWS wants to run agents, not just serve the models behind them.

    View source ↗
  6. 1d ago

    Better decisions at scale: How mathematical optimization delivers where intuition fails

    An explainer on where mathematical optimization fits in the AI landscape, with Innovation Center case studies. Thought-leadership rather than a capability change, off the main agentic thread.

    View source ↗