CloudflareA QUICker SASE client: re-building Proxy Mode
Re-architecting Cloudflare One proxy mode to run direct L4 proxying over QUIC (MASQUE) and HTTP/3, removing L3/WireGuard bottlenecks for dramatically faster, low-latency zero-trust traffic.
CloudflareHow Automatic Return Routing solves IP overlap
ARR enables zero-touch, stateful return routing to solve private IP overlap by remembering the originating tunnel for each flow within Cloudflare's Unified Routing, bypassing routing tables and returning traffic to its source site.
CloudflareEnding the "silent drop": how Dynamic Path MTU Discovery makes the Cloudflare One Client more resilient
Dynamic Path MTU Discovery enables the Cloudflare One Client to actively probe and optimize MTU with PMTUD (Path MTU Discovery) over MASQUE/QUIC, delivering resilient, uninterrupted connections across diverse networks.
AWS MLBuilding custom model provider for Strands Agents with LLMs hosted on SageMaker AI endpoints
Bridge SageMaker-hosted LLMs with Strands Agents by implementing a custom model parser that translates SageMaker responses into Bedrock Messages API format for seamless BYOC deployments.
AWS MLDrive organizational growth with Amazon Lex multi-developer CI/CD pipeline
Drive organizational growth by enabling a multi-developer Amazon Lex CI/CD pipeline with isolated environments, IaC-driven provisioning via AWS CDK, lexcli exports, Docker-based builds, automated testing, and staged deployments through GitLab CI/CD.
DatabricksMeet KARL: A Faster Agent for Enterprise Knowledge, powered by custom RL
KARL demonstrates a custom reinforcement learning approach to power a faster, cheaper enterprise knowledge agent with grounded reasoning, lower inference costs, and reduced latency.
Modular AIModular: Structured Mojo Kernels Part 1 - Peak Performance, Half the Code
Mojo enables Structured Mojo Kernels (composed of TileIO, TilePipeline, and TileOp) whose compile-time metaprogramming and RAII context managers decouple data movement, orchestration, and compute to deliver peak GPU performance with far less code and zero runtime overhead.
AWS MLUnlock powerful call center analytics with Amazon Nova foundation models
Explores powering call center analytics with Amazon Nova foundation models via Bedrock, delivering single-call and multi-call insights through Transcribe, Athena, S3, and a Streamlit UI, including sentiment analysis, protocol adherence, vulnerability detection, and AI-assisted BI queries.
AWS MLEmbed Amazon Quick Suite chat agents in enterprise applications
Embed Amazon Quick Suite chat agents directly into enterprise portals with a secure one-click deployment using the Quick Suite Embedding SDK, backed by CloudFront, Cognito, API Gateway, and Lambda, delivering JWT/OIDC-based authentication.
AWS MLHow Ricoh built a scalable intelligent document processing solution on AWS
Ricoh demonstrates scalable intelligent document processing on AWS by combining Amazon Textract, Amazon Bedrock foundation models, and the GenAI IDP Accelerator into a reusable, compliant, multi-tenant serverless framework that accelerates onboarding and increases throughput.
OpenAIHow Axios uses AI to help deliver high-impact local journalism
A technical overview of how Axios employs AI to automate and enhance the delivery of high-impact local journalism, from data ingestion to rapid reporting.
It Wasn’t a Culture Problem: Upleveling Alert Development at Airbnb
A technical exploration of how Airbnb upleveled alert development, arguing improvements come from tooling and practices rather than culture.