engblogs

A Rust-based exploration of ecdysis, a graceful-restart library that enables zero-downtime upgrades for Cloudflare's high-traffic services by forking and execing a new process while preserving live connections.

Examines transferring optimal global hyperparameters across model sizes through a unified Complete Parameterisation that scopes width, depth, batch size, and training duration, and demonstrates per-module hyperparameter transfer to speed up training of large language models.

Faster convergence in federated stochastic variational inequalities is achieved by refining Local Extra SGD guarantees and introducing the Local Inexact Proximal Point Algorithm with Extra Step (LIPPAX) to mitigate client drift and extend guarantees to composite VIs.

A practical guide to unifying Databricks and cloud infrastructure costs across multi-cloud environments via the Cloud Infra Cost Field Solution, delivering a single TCO view for governance, optimization, and informed decision-making.

Examines MIT's J-PAL Project AI Evidence, a funding-backed initiative to rigorously test AI solutions, generate evidence on what works, and responsibly scale the most effective innovations to fight poverty.

Lambda appoints Jerry Hunter as Vice Chairman for Compute Delivery and Special Advisor to guide industrial-scale AI infrastructure and long-term compute strategy, drawing on his AWS, Sun Microsystems, and Snap experience to accelerate hyperscale AI delivery.

Trace length emerges as a simple, practical uncertainty signal for large reasoning models, revealing how post-training reshapes confidence estimates, how it complements verbal confidence, and its implications for robust uncertainty quantification across tasks.

A study-driven roadmap mapping the UX design space for LLM-powered computer use agents, presenting a taxonomy of design factors and empirical insights to align agent behavior with diverse user needs and scenarios.

A practical guide to building robust, long-running MCP servers on Amazon Bedrock AgentCore with Strands Agents, enabling cross-session execution via persistent memory and asynchronous task management for seamless results and reliability.

Delivers a technical blueprint for an AI-powered talent acquisition system on AWS using Amazon Bedrock, knowledge bases, Lambda-driven workflows, and governance to boost efficiency while preserving human oversight.

A concise launch-focused overview of GPT-5.3-Codex-Spark and its implications for AI-powered code development.

Explores how low-bit inference and hardware-aware quantization—centered on MXFP-based mixed-precision—reduce memory and compute to enable fast, cost-effective AI on modern GPUs.