Building Trustworthy Knowledge Systems Through Ontology and Deterministic Reasoning
Why This Workshop
This workshop introduces the principles behind a hybrid knowledge system that combines LLM extraction with deterministic Datalog verification. The goal is to build the practical design skills needed to structure an organization's knowledge assets into a source-backed, ontology-based knowledge base (KB).
This workshop tackles LLM hallucination not through prompting, but through the knowledge system itself. We'll cover, from first principles, a hybrid architecture that combines LLMs, ontologies, and Datalog — where the LLM extracts knowledge from documents, a human reviews and approves it, and a deterministic Datalog reasoning engine verifies it for contradictions and policy violations. You'll experience this entire pipeline hands-on through a live demo of the open-source tool factlog. By the end, participants will have built a verifiable, source-backed knowledge base and learned the design patterns for extending it into an organization-wide ontology-based knowledge system.
The trust pipeline that runs through the entire workshop. Moving beyond the limits of chunk-based RAG, every answer is designed to carry both a source and a verification step.
What You'll Learn
Course Objectives
Understand the principles behind a hybrid knowledge system that combines LLM extraction with deterministic Datalog verification, and build the practical design skills to structure an organization's knowledge assets into an ontology-based knowledge base.
Course Summary
Covers the trust pipeline that runs from document → candidate facts → human review → accepted facts → engine verification → answer. You'll experience a verifiable knowledge system that catches contradictions and policy violations through Datalog logic checks, via a live demo (factlog) and hands-on practice.
- Hallucination and the limits of chunk-based RAG
- Ontology, knowledge graph, and Datalog fundamentals
- LLMs extract, deterministic engines decide — designing the trust boundary between "candidate" and "accepted"
- Demo: the full verification pipeline in the open-source tool factlog
- Practice the full flow: extraction → human review & approval → Datalog logic verification → source tracing
- Complete the entire process in the factlog hands-on environment
- Structuring reports, meeting notes, and internal wikis into a knowledge base
- Writing policy rules and detecting contradictions
- Open Q&A
- Roadmap and further learning resources
Bringing your own laptop is required to participate in the hands-on exercises. We recommend installing Python 3 and Claude Code in advance.
About the Speaker
Jeongseok Kim holds a PhD in Artificial Intelligence and brings over 20 years of experience in systems architecture design and project management. As CEO of Cleverplant, a company building neuro-symbolic AI decision-support systems, he leads the creation of business value through the convergence of ontologies and intelligent solutions. He has applied AI in real-world settings — ontology-based data analysis, standardization strategy, and open-source stack design — at global IT consulting firms and major corporations including SK Telecom and LG Electronics, and teaches practical ontology courses as an educator in the field of Knowledge Engineering.
Master Knowledge Systems Through Practice, Not Just Theory
This is a hands-on workshop with limited seats. Please bring a laptop and join us.
Register for the Workshop