engblogs

Databricks builds SAFE, an in-house feature flagging and experimentation platform, to decouple deployments from feature enablement and enable safe, scalable, low-latency rollouts across thousands of services while maintaining reliability.

A practical, hands-on guide to Gemini CLI—from installation to advanced workflows—covering context management with GEMINI.md, extensibility through Model Context Protocol (MCP) and extensions, and real-world use cases in software development, data analysis, and content creation.

Technical guide to architecting Amazon Bedrock AgentCore episodic memory, detailing episode extraction, memory storage, reflection, and benchmarks to enable agents to learn from past experiences across sessions.

A practical, one-line guide to building a production-grade Information Extraction workflow with Databricks Agent Bricks to convert FDA CRL PDFs into structured insights, emphasizing SME–engineer collaboration, formal evaluation, and scalable deployment.

Announcing BlackIce: a containerized, open-source red-teaming toolkit that bundles 14 AI security tools into a reproducible Docker-based environment for rapid prompt injection, data leakage, hallucination, and adversarial ML testing, with MITRE ATLAS and Databricks AI Security Framework alignment.

From ideation to global deployment, FOX—BASF’s AI-powered digital assistant for Finance and Controlling built on the Databricks platform and integrated with Teams—drives scalable, trusted insights and streamlined work across BASF.

Arctic Wolf's liquid clustering architecture delivers near real-time threat detection at petabyte scale by replacing partition-based layouts with workload-aware clustering, enabling fast access to recent data via Unity Catalog managed tables and a medallion data lake pipeline.

A technical look at how Higgsfield turns simple ideas into cinematic social videos, detailing scalable workflows, tooling choices, and production optimizations.

Showcases a scalable, modular multi-agent video processing workflow built with Strands Agents, Meta’s Llama 4, and Amazon Bedrock, where specialized agents are wrapped as tools to coordinate frame extraction, visual analysis, temporal reasoning, and summarization across AWS services.

bunq's Finn, an in-house generative AI assistant built on Amazon Bedrock, uses an orchestrator-based multi-agent architecture and agent-as-tool pattern to automate 97% of support across 38 languages with a 47-second average response.

RaMP incentives on Google Cloud accelerate AI-ready migration and modernization by offering platform credits, partner funds, and advanced-workload bonuses for SAP, Oracle, VMware, and data analytics.

Analyzes how AI, LLMs, and multimodal models will reshape quantitative investment management in 2026, emphasizing productivity gains, workflow improvements, and disciplined deployment across research and operations.