Skip to content

RESEARCH AND DEVELOPMENT

From research to product

Alternis, Notitia and Nova are not products first — they are the result of more than two decades of independent research. Since 2004, we have been investigating how artificial intelligence can genuinely support people, feeding every insight back into our technologies.

Independently verified

Our research and development expertise is not self-proclaimed. It has been audited by an independent body, and this work has produced our own research projects, whose results flow back into our products.

  • Our company was recognized for its independent research and development expertise.

  • We were a research partner in the renowned R&D project SPELL

  • We are an IHK-registered training company

What we research

  • Language research

    How machines can not only recognize spoken language but understand it — regardless of language, accent or background noise. The foundation of Alternis.

  • Knowledge representation

    How expertise can be modelled so that an AI applies it transparently, rather than merely guessing. The foundation of Notitia.

  • Human-AI collaboration

    How AI systems can be designed so that humans retain control and responsibility stays where it belongs. The foundation of Nova.

EnviroInfo 2026: Energy efficiency through hybrid AI

Presented as a full paper at the 40th International Symposium on Environmental Informatics (EnviroInfo 2026) in Łódź; the full version is to appear in the Springer book series “Progress in IS – Advances in Environmental Informatics”. Jason Rietzke (LiveReader GmbH) and Stefan Naumann (Trier University of Applied Sciences, Environmental Campus Birkenfeld) measured what it costs in energy to let a language model run a process that a rule-based system already knows.

For a realistic workflow they used the ANA emergency-call protocol from the SPELL research project — a semantic network of some 7,800 nodes. One identically scripted emergency call was run 30 times per strategy on the same hardware: once driven entirely by the language model, once with the model as a pure extractor, while Notitia's symbolic process control decides which question comes next. Energy consumption was measured per dialogue turn at the GPU's hardware counter, not estimated.

The hybrid approach took 71.5 seconds instead of 547.6 (factor 7.7) and 3.25 watt-hours instead of 28.92 (factor 8.9). The gain breaks down into two factors that multiply: 1.87 times fewer turns, because the rule set navigates deterministically, and 4.75 times cheaper turns, because the prompt needs no history and no state. The symbolic share itself carries no energetic weight — it comes in around 2,200 times below a single GPU step. Extrapolated to 1,000 emergency calls a day, that is some 2.3 tonnes of CO₂ equivalent a year, from an architectural decision alone.

Title slide of the talk "Efficiency Optimization Through Symbolic–Subsymbolic AI Integration" by Jason Rietzke and Stefan Naumann
"AI as a tool: a clear yes. AI as a replacement for one's own knowledge: a clear no." — Dr. Eric Rietzke, Managing Director

Publications

Our research is published in the open. All papers are freely accessible unless otherwise indicated.

  1. 2026

    Efficiency Optimization Through Symbolic–Subsymbolic AI Integration: An Empirical Energy and Performance Study

    Rietzke, J.; Naumann, S.

    Conference paperEnviroInfo 2026, Łódź · Springer: Progress in IS – Advances in Environmental Informatics Open Access

  2. 2025

    Hybrides wissensbasiertes Reasoning für wissensintensive Prozesse am Beispiel von Notrufabfragen

    Rietzke, E.; Maletzki, C.; Grumbach, L.; Bergmann, R.

    Book chapterHybride KI mit Machine Learning und Knowledge Graphs, Springer Vieweg, S. 297–318 Open Access

  3. 2024

    Towards Human-AI Interaction in Medical Emergency Call Handling

    Maletzki, C.; Elsenbast, C.; Reuter-Oppermann, M.

    Conference paper57th Hawaii International Conference on System Sciences (HICSS) Open Access

  4. Empowering Large Language Models in Hybrid Intelligence Systems Through Data-Centric Process Models

    Maletzki, C.; Rietzke, E.; Bergmann, R.

    Conference paperAAAI Spring Symposium MAKE 2024, Stanford University Open Access

  5. KI-gestützte Kommunikation in Notrufgesprächen (Alternis)

    Maletzki, C.; Rietzke, E.; Elsenbast, C.; Reuter-Oppermann, M.

    AbstractForum Rettungswissenschaften Open Access

  6. KI-gestützte Adaptive Notrufabfrage (KIANA)

    Maletzki, C.; Blaschke, F.; Hippler, B.

    AbstractForum Rettungswissenschaften Open Access

  7. 2023

    Towards Hybrid Intelligent Support Systems for Emergency Call Handling

    Maletzki, C.; Grumbach, L.; Rietzke, E.; Bergmann, R.

    Conference paperAAAI Spring Symposium MAKE 2023, San Francisco · CEUR Workshop Proceedings Open Access

  8. Interdisziplinäres Lagebild in Echtzeit – Grünbuch

    Zukunftsforum Öffentliche Sicherheit e.V. (ZOES); Rietzke, E. als Mitwirkender

    Book chapterZukunftsforum Öffentliche Sicherheit e.V. (ZOES) Open Access

  9. Implikationen von Mental Pictures in der Mensch-Computer Interaktion zur Entscheidungsunterstützung während der Notrufbearbeitung in Integrierten Leitstellen

    Elsenbast, C.; Maletzki, C.; Reuter-Oppermann, M.

    AbstractForum Rettungswissenschaften Open Access

  10. 2022

    SPELL – Intelligente Systeme zur Einsatzunterstützung

    Rietzke, E.; Elsenbast, C.

    AbstractForum Rettungswissenschaften Open Access

  11. Utilizing Expert Knowledge to Support Medical Emergency Call Handling

    Maletzki, C.; Rietzke, E.; Bergmann, R.

    Workshop paperFCR@KI 2022 (KI 2022) · CEUR Workshop Proceedings Open Access

  12. 2021

    Modellierung, Visualisierung und Ausführung wissensintensiver Prozesse unter Verwendung Semantischer Technologien

    Rietzke, E.

    PhD thesisUniversität Trier (OPUS) Open Access

  13. Execution of Knowledge-Intensive Processes by Utilizing Ontology-Based Reasoning (ODD-BP)

    Rietzke, E.; Maletzki, C.; Bergmann, R.; Kuhn, N.

    Journal articleJournal on Data Semantics, Vol. 10, S. 3–18 Open Access

  14. 2019

    ODD-BP – An Ontology- and Data-Driven Business Process Model

    Rietzke, E.; Bergmann, R.; Kuhn, N.

    Conference paperLWDA 2019 · CEUR Workshop Proceedings Open Access

  15. Utilizing Ontology-Based Reasoning to Support the Execution of Knowledge-Intensive Processes

    Maletzki, C.; Rietzke, E.; Grumbach, L.; Bergmann, R.; Kuhn, N.

    Conference paperBPM 2019, Workshop AI4BPM · Springer

  16. 2018

    SEMANAS – Semantic Support for Grant Application Processes

    Grumbach, L.; Rietzke, E.; Schwinn, M.; Bergmann, R. et al.

    Conference paperLWDA 2018 · CEUR Workshop Proceedings Open Access

  17. 2017

    Adaptive Business Process Visualization for a Data and Constraint-Based Workflow Approach

    Rietzke, E.; Bergmann, R.; Kuhn, N.

    Conference paperLWDA 2017 · CEUR Workshop Proceedings Open Access

  18. Semantically-Oriented Business Process Visualization for a Data and Constraint-Based Workflow Approach

    Rietzke, E.; Bergmann, R.; Kuhn, N.

    Conference paperBPM 2017, Workshop AI4BPM · Springer

  19. 2016

    SEMAFLEX – Semantic Integration of Flexible Workflow and Document Management

    Grumbach, L.; Rietzke, E.; Schwinn, M.; Bergmann, R.; Kuhn, N.

    Conference paperLWDA 2016 · CEUR Workshop Proceedings Open Access

Research with us

Are you working on similar questions, or considering how research and development could benefit your company? We welcome research partnerships and exchange with universities, institutes and other companies.