Research question
An AI team studies formal user intents. A research group compares matching approaches under uncertainty. A Semantic Web team investigates decentralized interoperability.
Research and open technical collaboration
Universal Market Matching Information Exchange
The most interesting collaborations begin with an open question.
UMMINEX is a published open technical foundation. We are looking for people and organizations that want to research, test, or implement the fundamental questions around it independently.
Nobody needs to adopt UMMINEX operationally or commit to it as a future architecture. Independent research, prototypes, alternative implementations, benchmarks, criticism, and demonstrators are valuable outcomes.
Collaboration
Open questions that reach beyond UMMINEX.
UMMINEX touches research and engineering questions that go far beyond the project itself: How can human intent be represented formally and verifiably? How can independent systems compare complex offers and requests? How can AI agents communicate market intents without proprietary interfaces? And how can different semantic worlds be connected without centralizing them?
Test case
Tourism, procurement, and other domains make these questions concrete. They can serve as demonstrators, benchmarks, or experimental environments.
You do not need to adopt UMMINEX. Bring an open question, a technical approach, or an interesting counterexample.
What we bring
A starting point, not just an idea.
UMMINEX already comes with a specification, implementation knowledge, extensive tests, and documented architectural decisions. factsoft AG can contribute this knowledge and help scope a suitable project. A partner does not need to start from zero.
The actual development or research does not have to be carried out by factsoft AG.
Open challenges
Six directions for independent work
These examples are not a roadmap. They are research, engineering, and demonstrator questions that can be investigated independently.
01
Translate natural language reliably into UMMINEX
People describe market intentions in ordinary language. The central question is whether a formal representation can later be checked to confirm that it still expresses what the person meant.
Especially relevant are reconstructing context, genuine ambiguity, information that may remain UNKNOWN, asking only necessary questions, formal representation, deterministic validation, meaning reconstruction, and measurable evaluation.
Suitable for: NLP and LLM research, formal semantics, AI labs, universities, and agent-system developers.
02
An independent UMMINEX matcher
An independent reference or research implementation can explore constraints, resources, alternatives, multiple parties, time, quantities, uncertainty, and progressive resolution.
UMMINEX already has matching components. Another implementation can independently reproduce, compare, benchmark, or criticize approaches.
Suitable for: operations research, constraint programming, universities, optimization, and marketplace or search research.
03
An open market language for AI agents
An AI agent may understand what its user wants in a conversation. The difficulty begins when it must communicate that intent to independent providers, data sources, or other agents instead of translating it anew for every proprietary interface.
UMMINEX could be a formal layer: human intent → AI agent → formal Market Intent → independent providers, data sources, or other agents → matching result. This is an open research and application direction, not an adopted architecture.
Suitable for: agent systems, LLM research, multi-agent interoperability, universities, and search technology.
04
Connect different data worlds
Industries and providers use different schemas, vocabularies, classifications, taxonomies, and attributes.
Possible work includes mappings, equivalence relationships, broader and narrower concepts, semantic resolution, public vocabularies, and web standards – without centralizing vocabularies.
Suitable for: Semantic Web, Knowledge Graphs, semantic interoperability, PIM/MDM, data integration, and open standards.
05
Procurement as a matching demonstrator
A company needs 800–1,000 aluminium housings per month from March, delivered to a defined location with technical and commercial conditions. A supplier describes what it can deliver.
This case can test whether independent supply and demand descriptions can be represented and compared formally without forcing both sides into the same marketplace database.
Suitable for: procurement research, industry, ERP, independent implementations, and benchmarks.
06
Decentralized availability as a tourism demonstrator
Hotels and accommodation providers could publish offers and availability in machine-readable form. A local tourism organization could show available accommodation without requiring every provider to maintain inventory in a new central marketplace.
This is a demonstrator for decentralized publication and discovery, not an existing production deployment.
Suitable for: tourism research, municipalities, hotel software, accommodation providers, universities, and independent prototypes.
Open collaboration
Who can work independently
Research and teaching
Universities, research institutes, and groups for NLP, LLMs, formal semantics, and AI agents.
Matching and interoperability
Operations research, constraint programming, Semantic Web, Knowledge Graphs, and open protocols.
Independent development
Open-source developers, independent implementers, and teams for benchmarks, experiments, or prototypes.
Research-oriented companies
Companies with their own R&D or fundamental interoperability and matching questions.
Funded research, joint proposals, independently funded work, and clearly scoped open technical projects are welcome. Funding is not required.
Starting point
What already exists
Partners do not start from zero. The public material includes formal semantics, specification, validation and conformance work, test material, and a repository released under the Apache License 2.0.
UMMINEX on GitHub: source and specification
The technical provenance of the public release is documented.
Governance
Openly licensed, not intended to depend permanently on one company
UMMINEX was originally developed by factsoft AG. The published material is openly available under the Apache License 2.0. factsoft AG currently carries organizational responsibility for the present project phase.
In the long term, however, UMMINEX should not depend permanently on a single company. If several independent implementers and users emerge, an appropriate independent stewardship or governance model can be established together. Such a model does not exist today.
Contact
Bring a question, a counterexample, or a test domain.
You do not need a finished project plan. Perhaps you study semantic preservation in LLM-generated formal representations, multi-agent interoperability, constraint solving with incomplete information, or semantic matching. Perhaps you want to build an independent implementation or demonstrator.
Or you have a concrete application question. Then we can explore together whether UMMINEX is suitable. Maybe it fits. Maybe it does not. That is exactly what we want to find out.
The public material is available at github.com/umminex/umminex.