← All speakers

Bio, Work & Ideas

Julián Duque

Conference affiliation: Heroku · 2025

On this page

Julián Duque is a principal developer advocate at Heroku, a Node.js educator, and an organizer of JSConf Colombia and NodeConf Colombia. His work brings the hard-won lessons of production JavaScript—reliable infrastructure, clear operational boundaries, and accessible developer tooling—to AI applications.

From community organizing to applied AI

Duque helped launch JSConf Colombia in 2013 and later organized NodeConf Colombia, whose inaugural 2019 gathering featured practical workshops and scholarships. His community work also included MedellinJS and the broader Node.js ecosystem. His account of building Colombian developer communities emphasizes making technical education and professional networks more accessible.

Before joining Heroku, Duque worked at NodeSource and advised companies as a Node.js solutions architect. His writing on production error handling advocates useful logs, visible failures, graceful shutdowns, and predictable recovery. At Heroku, he expanded into technical writing, workshops, cloud infrastructure, and AI-assisted development. In 2024, he published guides to connecting ChatGPT functions with deployed Node.js applications and building a custom GPT backed by a Heroku API.

Practical ideas and projects

  • Explicit permissions for AI agents. In a workshop co-led with Anush DSouza, Duque showed how short-lived Heroku dynos can execute tools while applications receive explicitly scoped access to compute and data. The practical question is what an agent can run, where it runs, and which resources it can reach.
  • Schema-aware, read-only database access. Duque demonstrated an agent inspecting a PostgreSQL schema before generating SQL and querying a read-only follower. This grounds generated queries in actual database structure while limiting the consequences of unsafe model behavior.
  • Model Context Protocol integrations. His account of using Cursor with Heroku MCP describes retrieving logs, managing deployments, consulting documentation, and incorporating design context directly within an editor. His workshop also explored custom MCP servers and authenticated remote access.
  • Open integrations for familiar developer tools. His projects include heroku-langchain, connecting LangChainJS with Heroku inference services; heroku-ai-provider, a provider for the Vercel AI SDK; and n8n-nodes-heroku-ai, which brings Heroku AI into n8n automation workflows. His Nodebooks project extends his commitment to hands-on education through interactive Node.js notebooks.

Read the topics behind these talks

1 conference talk

References