engblogs

Discover how Serverless JARs enable fast-starting, versionless Spark jobs on fully managed Databricks serverless compute, with no infrastructure to manage and pay-for-what-you-use, while development leverages Databricks Connect and deployment via Asset Bundles.

Ship production-grade, governance-backed enterprise AI agents to business users by stitching together Agent Bricks Knowledge Assistant, Databricks Apps, and Databricks One, with citation-backed responses and robust security.

GCI partners with Databricks to transform vast Alaskan data into governed, AI-assisted analytics and actionable dashboards—leveraging Unity Catalog, Metric Views, AI/BI Genie, and Dynamic Dashboards to improve customer experience and guide network investment.

Automates Bedrock model availability and cross-region routing for Agentforce across 30+ AWS accounts, slashing discovery from days to minutes with Lambda-based data collection, CloudWatch metrics, and Grafana-driven observability to improve latency, resilience, and compliance.

Explores deploying disaggregated inference on AWS with llm-d, detailing prefill/decode separation, cache-aware routing, MoE and tiered KV caching, and deployment on SageMaker HyperPod and EKS.

Guidance-driven blueprint for deploying agentic AI in enterprises, mapping leadership personas to governance, measurement, and workflow orchestration to turn autonomous agents into scalable, auditable business capability.

A technical overview of Google Cloud and NVIDIA's co-engineered AI Hypercomputer stack that accelerates agentic AI, large-model inference, and industry-scale workloads through G4 VMs, Vera Rubin NVL72 integration, Dynamo-enabled orchestration, and Vertex AI enhancements.

BigQuery Studio's Gemini-powered assistant delivers context-aware interoperability, intelligent resource discovery across projects, and instant job analysis to turn analytics overhead into proactive AI-assisted data workflows.

Modular's NVIDIA GTC 2026 showcase highlights MAX on Blackwell, a Mojo-based port of CUTLASS kernels to Mojo with 130.7 TFLOPS on B200, and DeepSeek V3 running in Modular Cloud, illustrating end-to-end diffusion pipelines, kernel optimization, and scalable inference.

Analyzes the 2025 ransomware landscape—TTP evolution, data-theft extortion trends, and shifts in the RaaS ecosystem, highlighting cross-platform deployments and implications for defense and recovery.

An end-to-end guide to building a scalable, governed offline feature store with SageMaker Unified Studio and SageMaker Catalog, using Apache Iceberg-backed S3 Tables, Lake Formation access control, and a publish-subscribe workflow to enable cross-team collaboration and reproducible ML features.

Workhuman demonstrates a scalable approach to multi-tenant, self-service analytics by embedding Amazon QuickSight dashboards, enforcing strict data isolation with namespaces and row-level security, and automating deployment through a CI/CD pipeline to deliver customized insights across SaaS customers.