engblogs

A concise technical overview of the cross-sector partnership between OpenAI, SoftBank Group, and SB Energy to explore AI-powered energy technologies.

Replace read-only ephemeral sandboxes with Sprites—durable, instantly bootable computers that support checkpoint/restore, ample storage, and global Anycast access for seamless development-to-prod workflows.

An end-to-end guide to accelerating LLM inference via post-training weight-and-activation quantization using AWQ and GPTQ on Amazon SageMaker AI, reducing memory and latency without retraining.

Introducing the Snorkel Agentic Coding Benchmark—a real-world, end-to-end evaluation suite for AI coding agents that spans multiple languages, 100 multi-step tasks across four difficulty tiers, and rigorous long-horizon planning, error recovery, and sandboxed execution.

A principled taxonomy for evaluating speech models that maps evaluation types to model capabilities and task requirements, revealing gaps in prosody, interaction, and reasoning to guide benchmark design and practical model selection.

Case study: Hackensack Meridian Health de-risks a large Google Cloud network migration by using VPC Flow Logs and Flow Analyzer to visualize Cloud Interconnect traffic from VLAN attachments via Sankey diagrams.

Introducing PPGen, a biophysical model that maps photoplethysmography signals to interpretable optical and physiological tissue parameters, and Hybrid Amortized Inference (HAI) for fast, robust, and scalable parameter estimation in wearables, balancing deep learning performance with clinical interpretability and hardware-aware design.

A memory-augmented pretraining approach using hierarchical memories to separate long-tail knowledge from common knowledge, with a small anchor model accessing a large memory bank for memory-efficient inference and fine-tuning on resource-constrained devices.

Three-stage, domain-adaptive YOLOv11 PPE detection trained on TrueLook data, deployed on SageMaker AI via automated Pipelines and Model Registry to deliver real-time construction-site safety monitoring with governance and continuous improvement.

Explores end-to-end sentiment analysis for text and audio using AWS Generative AI services, detailing architectures, experiments, challenges, and future directions.

Beekeeper builds a Bedrock-powered, feedback-driven leaderboard to optimize LLM prompts and personalization for frontline workers, enabling dynamic model/prompt selection and tenant-specific customization.

Datadog leverages Codex to perform system-level code reviews, illustrating how AI-assisted analysis enhances code quality and reliability across complex, distributed systems.