What happens in the milliseconds after you tap pay
Summary
This sample Databricks App demonstrates how to achieve low-latency real-time fraud scoring by pairing route-optimized Model Serving with Lakebase Postgres for online feature lookups. Under load testing of 5,000 requests, this architecture achieved end-to-end latencies of 27 ms at p50 and 37 ms at p95 while maintaining a 100% success rate.
Summary generated by brickster.ai. For the full article, follow the source link above.
More from Databricks Blog
Your AI is ready. Your data foundation probably isn’t
Cushman & Wakefield successfully built its enterprise AI core by embedding technologists into business units under a product operating model, prioritizing trust and a co-created capital investment model over short-term AI pilots. Using Databricks and Genie, the company matured its operating model three times in four years and accelerated its idea-to-outcome timelines from months to days.
From experiment to insight: how Dotmatics Luma and Databricks make AI-ready science a reality
Dotmatics Luma and Databricks have partnered to deliver a unified data foundation that combines scientific context and instrument connectivity with enterprise-scale storage, governance, and AI tooling. This collaboration enables life sciences organizations to overcome data fragmentation and build a repeatable, trustworthy foundation for AI-ready science.
Unified context: The missing layer for enterprise AI coworkers
Enterprise AI coworkers require a unified context layer, powered by Genie Ontology, to bridge the gap between scattered data and decision-ready business insights. This foundation enables Genie One to operate as a trusted AI coworker that can explain its reasoning on governed data and turn decisions into direct action across tools like Slack, Teams, and dashboards.
The skills gap behind agentic AI — and how Databricks is closing it with a new context engineer certification and agent trainings
Databricks has launched an industry-first Context Engineer Associate certification and expanded its learning catalog with targeted courses to help practitioners build reliable agentic AI systems. The platform is also the first to publish official guidance on utilizing agents like ChatGPT, Genie, and Claude to help candidates effectively study for these certification exams.
Introducing Apache Spark 4.2
Apache Spark 4.2 introduces governed business definitions via metric views, AI-native analytics features like vector retrieval, and simplified real-time data processing through Auto CDC and Real-Time Mode. This release also expands Spark's accessibility from external services and AI agents by leveraging Spark Connect, Arrow-first Python execution, and Python Data Sources.
Inkling model from Thinking Machines Lab now on Databricks
The open-weights Inkling model from Thinking Machines Lab is now available on Databricks through the Unity AI Gateway, allowing practitioners to easily build and deploy AI agents and coding applications on enterprise data. This release enables teams to leverage a model optimized for coding, agentic reasoning, and multimodal inputs while maintaining centralized security, cost controls, and observability.