ANN, RAG, MCP, or Agentic AI – It’s All About Data, and Mostly Open Source
As AI application architectures rapidly evolve—from basic RAG pipelines to autonomous Agentic AI and standard context protocols like MCP—the true center of gravity remains constant: data. While proprietary foundation models dominate headlines, open source software forms the core infrastructure powering data integration, vector retrieval, and agent orchestration.
This talk explores the vital role of open source across the modern AI stack. We will trace the architectural evolution connecting retrieval mechanisms to model context protocols and multi-agent workflows, unpack the critical distinctions between hosted LLMs, open-weights, and true open-source models, and demonstrate how open-source data tooling ensures innovation, governance, and sovereignty in enterprise AI deployments.
Key Takeaways for Attendees:
A pragmatic architecture breakdown of modern AI data stacks (RAG, MCP, and Agentic frameworks).
Guidelines on evaluating Open Source vs. Open Weight vs. Hosted API models for privacy and compliance.
Architectural patterns using open-source tools to maintain innovation and sovereignty.
Marc Linster is an open-source technology strategist and an instructor at the Digital Learning Hub in Luxembourg. With decades of experience across the US, Asia, and Europe, Marc was formerly CTO and Chief Product Officer at EnterpriseDB (EDB), guiding global enterprises through high-stakes migrations to open source.
He is the co-author of “AI-Ready PostgreSQL 18”, holds a Ph.D. in Computer Science, and is an active member of the Luxembourg Business Angel Network.
Marc is a co-organizer of PGConf.EU, the largest annual Postgres conference, and he is actively advising leading PostgreSQL companies, such as DBtune, on using AI and Machine Learning to create high-performance autonomous database solutions.