A Medium write-up describes building an agentic harness that generates Databricks medallion-architecture pipelines. It highlights a cost concern for agentic loops: repeated prompting of the same context drives up API charges, with Databricks Genie billed per interaction at roughly eleven cents a question.
A write-up on applying Databricks Genie, an LLM-powered data querying assistant, to an insurer's private debt portfolio monitoring workflow. It covers moving from quarterly portfolio files to weekly compliance checks against investment guidelines for committee decisions. The audience is investment management teams considering LLM-based analytics on governed data platforms.
A write-up on applying Databricks Genie, an LLM-powered data querying assistant, to an insurer's private debt portfolio monitoring workflow. It covers moving from quarterly portfolio files to weekly compliance checks against investment guidelines for committee decisions. The audience is investment management teams considering LLM-based analytics on governed data platforms.
A pilot of Databricks Genie agent with merchant and finance teams found that the conversational BI tool achieves usable accuracy only when scoped to a narrow, well-modeled domain (single star schema, ~15 tables) with sample questions and verified SQL as instructions. Pointing it at a broad, messy schema resulted in poor performance.
A pilot of Databricks Genie agent with merchant and finance teams found that the conversational BI tool achieves usable accuracy only when scoped to a narrow, well-modeled domain (single star schema, ~15 tables) with sample questions and verified SQL as instructions. Pointing it at a broad, messy schema resulted in poor performance.
Databricks has moved away from betting on a single agent framework and instead built a harness, called Omnigent, to support all of them. This allows developers to use multiple agent frameworks within Databricks' platform.
Databricks has moved away from betting on a single agent framework and instead built a harness, called Omnigent, to support all of them. This allows developers to use multiple agent frameworks within Databricks' platform.
A Medium article discusses the concept of stackable traits for agent prompts, suggesting that the real challenge lies not in writing prompts but in composing them effectively. The post highlights a modular approach to prompt engineering.
A Medium article discusses the concept of stackable traits for agent prompts, suggesting that the real challenge lies not in writing prompts but in composing them effectively. The post highlights a modular approach to prompt engineering.
A Medium article describes the agent harnesses that power Microsoft's AI ecosystem, moving beyond simple prompting to orchestrate agents. The piece explains the underlying infrastructure and design patterns used to build and manage AI agents at scale.
A Medium article describes the agent harnesses that power Microsoft's AI ecosystem, moving beyond simple prompting to orchestrate agents. The piece explains the underlying infrastructure and design patterns used to build and manage AI agents at scale.