Research
Medical Knowledge & AI Research
Clinical NLP and knowledge representation work focused on turning guideline knowledge into structured, traceable and verifiable forms.
Explore↗
Complex systems · Applied AI · Decision support
Julia
Kewitz
Turning complex knowledge into systems people can understand, challenge, and use.
I work at the intersection of complex systems, applied AI, and decision support. My background spans more than fifteen years in transportation, logistics, and operational systems, while my current research explores how expert knowledge can be transformed into machine-usable forms without losing the context, evidence, and conditions that give it meaning.
contact@gnodus.ai
Current focus
How expert knowledge can be made usable by AI without losing what makes it reliable.
JK
01 · About
From complex systems to computable knowledge.
I have spent more than fifteen years working with complex operational and decision-support problems across transportation and logistics.
Across very different projects, I kept encountering a similar problem: important knowledge is fragmented across data, documents, rules, and people. Much of my work has involved making that complexity explicit — translating ambiguous requirements, constraints, and expert knowledge into analytical models, system logic, and decision-support structures.
Over time, this led me deeper into applied mathematics, data science, and AI. During my Master’s degree in Applied Mathematics and Informatics at HSE University, I developed a system for transforming oncology clinical guidelines into structured, source-linked knowledge. That work raised a broader question that now drives my research: what gets lost when expert knowledge is transformed into something a computer can use?
I am now exploring how medical knowledge can remain meaningful, traceable, and verifiable as it moves through extraction, representation, retrieval, and use by AI systems.
In parallel, I am building Gnodus, an early-stage research-driven company exploring whether this same problem can become the basis for a useful technology and business.
02 · Selected work
Four areas that connect my research, technical practice, and real-world systems work.
A curated view of the work that best explains how I think — with direct links to selected repositories where they add useful evidence.
Technical practice
Language, Retrieval & Representation
Experiments in language modelling, dense retrieval, semantic representation and multimodal search.
Explore↗
Applied systems
Decision Systems & Applied ML
Analytical systems where models, business logic and operational constraints support real decisions in risk, pricing, auctions and planning.
Explore↗
Building
Gnodus
An early-stage research-driven company exploring how expert knowledge can be made reliably usable by AI — and where that capability can become a real product.
Explore↗
03 · Research
What gets lost when expert knowledge becomes machine-usable?
Current question
How can expert knowledge become machine-usable without losing context, provenance, and meaning?
Clinical guidelines are the current research setting: a difficult, high-stakes domain where trustworthy knowledge representation matters.
01 · PreserveWhat information must travel with a knowledge claim?
02 · RepresentHow should context, evidence and applicability be encoded?
03 · RetrieveHow can systems recover the right knowledge together with its qualifications?
04 · EvaluateHow do we measure knowledge quality beyond extraction accuracy?
04 · Building
Gnodus
A separate entity, but part of the same intellectual trajectory.
Early-stage research-driven venture
From information access to knowledge quality.
Gnodus is an early-stage company being built around a simple problem: giving AI access to information is not enough if the context, evidence, and conditions that make that information reliable are lost along the way.
Medicine is the first research and application domain. The current work focuses on understanding where this problem creates real value for expert users and what part of it can become a reusable technology.