engblogs

Wrap-up of the Mila World Modeling Workshop detailing advances in perceptual world models, multimodal representation, scalable architectures, and pathways to safe, autonomous AI, with highlights on latent particle models and JEPA approaches discussed by leading researchers.

Structured Mojo Kernels demonstrate a separation-of-concerns GPU kernel architecture in Mojo that delivers fast, maintainable, and portable code through compile-time metaprogramming and zero-cost abstractions, illustrated by TileIO, TilePipeline, and TileOp.

A concise technical update detailing progress, architecture decisions, and impact assessment of our mental health-related initiatives.

Explores the implications of OpenAI and Amazon's strategic partnership for AI tooling, cloud services, and enterprise deployment.

A practical blueprint for scaling AI to everyone, detailing actionable patterns, tooling, and governance for accessible and responsible deployment.

Explores a stateful runtime environment for agents within Amazon Bedrock, enabling persistent agent state, lifecycle management, and resilient execution in the cloud.

A concise overview of the joint statement from OpenAI and Microsoft, highlighting the collaboration and its strategic implications for AI development.

A concise technical guide to diagnosing and bridging online–offline discrepancies in Pinterest’s L1 conversion models, with a focus on signal alignment and attribution accuracy.

Case study of developing and field-testing an underwater navigation algorithm that enables a human diver and an autonomous vehicle to collaborate in GPS-denied environments.

Consolidate scattered APIs into a single control plane with API Hub and API Gateway, while using the specification boost add-on to make APIs AI-ready and easily discoverable by agents.

Design a scalable agentic chatbot memory architecture using polyglot storage across Memorystore for Redis, Cloud Bigtable, and BigQuery to deliver fast real-time context with durable long-term history and analytics-ready insights.

Google Cloud's Cloud CISO Perspectives reveals how Google tackles foundational security, governance, AI risk, and agentic defense—advancing security operations, threat intelligence, and resilient strategies in an AI-driven era.