Background
Complex systems
15+ years across transportation, logistics, operations, planning and decision-support environments.
Personal site prototype · v0.1
Julia
Kewitz
Designing systems that turn knowledge into decisions.
I work with complex systems, data and expert knowledge. My background spans transportation, decision support and applied analytics. I currently explore NLP and computational knowledge representation while building Gnodus.
Current focus
Computable expert knowledge, NLP, semantic alignment and verifiable decision support.
JK
01 · About
One trajectory, several domains.
The site does not force one job title. It makes the trajectory understandable: real-world systems → analytics → computational methods → knowledge systems.
How I work
Structure before solutions
I like turning messy systems into explicit models, data structures and decision logic people can interrogate.
Direction
Knowledge systems
Current research explores how expert knowledge can remain contextual, verifiable and useful after becoming computational.
02 · Selected work
Evidence, not a list of skills.
Click any card. The prototype opens a floating detail window instead of sending you away from the page.
Professional work
Transportation Decision Support
Planning, operations and decision-support work across complex transport systems.
Explore↗
Applied data engineering
NorthCart Feature Pipeline
A reproducible batch pipeline generating customer-level features while preventing future-data leakage.
Explore↗
Applied machine learning
Demand Forecasting with ML
End-to-end ML workflow with feature engineering, stability monitoring and model evaluation.
Explore↗
Independent research
Clinical Knowledge Representation
Experiments in extracting, aligning and representing clinical knowledge for verifiable decision support.
Explore↗
03 · Research
A research question instead of buzzwords.
Current question
How can expert knowledge become computable without losing context, provenance and meaning?
Clinical guidelines are the current research setting: a difficult, high-stakes domain where trustworthy knowledge representation matters.
01 · ExtractionClinical entities, recommendations and contextual signals.
02 · AlignmentNormalizing concepts across heterogeneous sources and terminologies.
03 · RepresentationStructures that preserve relationships, evidence and provenance.
04 · VerificationMaking outputs inspectable by experts instead of opaque.
04 · Building
Gnodus
A separate entity, but part of the same intellectual trajectory.
gnodus
Research-driven initiative
Verifiable knowledge systems for expert domains.
Gnodus explores how heterogeneous expert knowledge can be transformed into transparent, inspectable decision-support systems. The current research laboratory is clinical knowledge.
05 · Writing & talks
A public record of ideas over time.
This can start small and later absorb publications, MBA writing, talks, research notes and conference appearances.
06 · Currently curious about
A little less CV, a little more human.
Computable knowledgeClinical decision systemsMathematical thinkingIndustrial AIResearch methods