From SDKs to Agents: Building with R and Python on Databricks
summary
Demo shows a working pipeline that lets Positron Assistant call a simple R MCP server (about 20 lines for the demo, full implementation ~150 lines) which uses brickster and ellmer to query Databricks, run dplyr/PySpark, and push a Confluence report. Access uses OAuth so Unity Catalog permissions are enforced, compute uses Databricks serverless autoscaling, and LLM calls go through the AI gateway (Anthropic Claude used in the demo) with token and cost guardrails.
AI-written summary · about
chapters (7)
- 0:00Building with R and Python on Databricks
- 1:03Automating reports with Positron and MCP
- 4:16MCP server connects Databricks and Positron
- 8:21Running MCP server in Positron
- 10:32Security and compute on Databricks
- 13:23Key takeaways and try it
- 14:50Q&A
related talks
18mElevating enterprise data through open source LLMs
33mBuild your own advanced AI solutions on Azure Databricks
19mBuilding Agents at Cloud Scale
36mDatabricks as the Foundation for GenAI: What’s New and What Works
17mHow will AI transform your data platform?
52mSerg Masís | AI Track | Compass AI & Data Summit 2025
40mBuilding Agents with Agent Bricks and MCP
45mManaging Data & Analytics Projects Effectively