engblogs

Scale AI in South Africa by leveraging Amazon Bedrock's global cross-Region inference to run Claude 4.5 models from the Cape Town region, routing requests to capacity across Regions while maintaining data residency and centralized logging.

Demonstrates how to simplify ModelOps by using Amazon SageMaker AI Projects with S3-based templates to provision secure, versioned environments and GitHub Actions powered CI/CD.

An end-to-end guide to evaluating generative AI models using Amazon Nova LLM-as-a-Judge on SageMaker AI, enabling scalable, impartial pairwise comparisons and rich metrics such as win rates and confidence intervals across real-world benchmarks.

A pragmatic framework for implementing AI governance that preserves speed and innovation while building trust, using observability and lifecycle discipline to govern data, models, and agents.

Building vertical microfrontends on Cloudflare’s platform with a Router Worker to stitch independent Workers into a single domain, enabling end-to-end ownership per URL path via service bindings, HTML rewriting, and CSS view transitions for a seamless user experience.

Unpacks the hidden technical debt in GenAI systems by contrasting classical ML workflows with GenAI execution and outlining governance, observability, prompting, and tooling practices to manage traces, tool sprawl, and human feedback.

Modular 26.1 advances programmable, portable AI infrastructure by introducing the MAX Python API, strengthening Mojo, and expanding Apple Silicon GPU support to streamline building, debugging, and deploying GenAI models across heterogeneous hardware.

This post examines how Taisei Corporation leverages ChatGPT to shape the next generation of talent through AI-powered training and scalable upskilling.

A practical blueprint for scaling enterprise content review with a three-agent workflow (scanner, verification, and recommendation) powered by Strands Agents and Amazon Bedrock AgentCore to keep technical content accurate and up-to-date.

Modular 26.1 introduces the MAX Python API and Mojo language enhancements to simplify building and deploying GenAI models across heterogeneous hardware, improve developer velocity, and expand Apple Silicon GPU support for faster, portable AI inference.

Details the deprecation of GPT-4o, GPT-4.1, GPT-4.1 mini, and OpenAI o4-mini in ChatGPT and the practical implications for integrations.

Explores the architecture, governance, and operation of OpenAI’s internal data agent, highlighting how autonomous data pipelines enable AI-driven workflows.