What we study, and what it changes in practice.
Our research in AI, from funded projects to publications and talks. See what left the lab.
Research
- Research contributions
- 21
- Certified R&D projects
- 3
- Conferences and venues
- 8
- Research partners
- 9
Three research areas, one direction.
Everything we publish feeds the AI native transformation of companies.
01
SMB Research
We study how successful mid-sized companies stay resilient and grow their people, and bring those lessons into every client project.
02
Knowledge Management
We research how a company’s own knowledge becomes something AI can answer from, with sources, and how it stays when people leave.
03
Agentic Coding
We test what coding agents need to remember, so our clients’ teams can rely on them in production.
Publication · 11 to 12 August 2026
Michael Banf et al.FGWM 2026 at KI 2026, Bremen · position paper
From documentation to knowledge vessels: agentic AI for tacit knowledge in manufacturing.
Michael Banf et al.FGWM 2026 at KI 2026, Bremen · position paperAll research
- Sep 2026A graph-based competency representation system with structure-preserving retrieval and evidence-based BARS scoring for AI-supported hiringPerelyn GmbH · Certified R&D project (BSFZ), German research allowanceCertificationCompetency geometry
- Sep 2026Non-Euclidean competency geometry through product-manifold embeddings and calibrated LLM scoring for aptitude diagnosticsPerelyn GmbH · Certified R&D project (BSFZ), German research allowanceCertificationCompetency geometry
- Aug 2026Memory-Aware Software Engineering Agents: Architectures, Mechanisms, and Open GapsMichael Banf et al. · AI4SE workshop at KI 2026, Bremen, organized by Fraunhofer IESEPublicationAgent memory
- Aug 2026From Documentation to Knowledge Vessels: Agentic AI for Tacit Knowledge Preservation in ManufacturingMichael Banf et al. · FGWM 2026 at KI 2026, Bremen · position paperPublicationAI in production
- Aug 2026Graph-Structured Memory for Cognitive Agents: From Knowledge Graphs to Non-Euclidean GeometryMichael Banf et al. · FGWM 2026 at KI 2026, BremenPublicationAgent memory
- Aug 2026Hyperbolic Embeddings for Competency Taxonomies: Hierarchy-Preserving Retrieval on a Real-World Skill GraphMichael Banf and Johannes Kuhn · FGWM 2026 at KI 2026, BremenPublicationCompetency geometry
- Jun 2026Curvature-guided Graph-to-Hypergraph Structural Lifting Alleviates Over-Squashing in Graph LearningMichael Banf · IOCMA 2026, 2nd International Online Conference on Mathematics and Applications (MDPI) · posterPublicationTopological deep learning
- Jun 2026Decoding Values: Measuring Culture Embedded in Large Language ModelsAnton Steuer et al. · Helmholtz AI Conference 2026, Munich · posterPublicationValues in LLMs
- Jan 2026A tripartite knowledge graph with topological lifting for concept-anchored answers to expert questionsPerelyn GmbH · Certified R&D project (BSFZ), German research allowanceCertificationKnowledge graphs
- 2026R&D seal of the BSFZ, 2026Perelyn GmbH · Bescheinigungsstelle Forschungszulage (BSFZ), the federal certification body for the German research allowanceCertificationCertification
- Dec 2025Topological Deep Learning Challenge 2025: Expanding the Data LandscapeBernárdez et al. · TAG-DS 2025, UC San Diego · PMLR 321ChallengeTopological deep learning
- Nov 2025Building Evolving AI Agents via Dynamic Memory Representations Using Temporal Knowledge GraphsMichael Banf · NODES 2025, Neo4j online developer conferenceTalkAgent memory
- Sep 2025Model Context Protocol in Manufacturing: A Comprehensive Framework for AI-Driven Industry 4.0?Michael Banf and Johannes Kuhn · 2nd Workshop on AI in Production (AIP) at KI 2025, Potsdam · proceedings publishedPublicationAI in production
- Sep 2025Tripartite-GraphRAG via Plugin OntologiesMichael Banf and Johannes Kuhn · FGWM workshop at KI 2025, Potsdam (GI Special Interest Group on Knowledge Management)PublicationKnowledge graphs
- Aug 2025A Remedy for Over-Squashing in Graph Learning via Forman-Ricci Curvature based Graph-to-Hypergraph Structural LiftingMichael Banf et al. · arXiv:2508.11390, preprintPublicationTopological deep learning
- Aug 2025HOOC 2025: Higher Order Opportunities and ChallengesPerelyn research group · Workshop on higher-order networks, RWTH AachenTalkTopological deep learning
- Aug 2025TopoBench: A Framework for Benchmarking Topological Deep LearningTelyatnikov et al. · Journal of Data-centric Machine Learning Research (DMLR)PublicationTopological deep learning
- Jun 2025A Topological Remedy for Over-Squashing in Graph Learning via Forman-Ricci Curvature based Graph-to-Hypergraph Structural LiftingMichael Banf et al. · Helmholtz AI Conference 2025, Karlsruhe · posterPublicationTopological deep learning
- 2024ICML Topological Deep Learning Challenge 2024: Beyond the Graph DomainBernárdez et al. · ICML 2024, GRaM workshop · PMLRChallengeTopological deep learning
- Jul 2023Deutschlands Mittelstand #3: Innovationsmotor PersonalMichael Banf and RKW Kompetenzzentrum · Study series of the RKW Kompetenzzentrum, Eschborn · data: Oskar-Patzelt-Stiftung, Großer Preis des Mittelstandes 2022PublicationSMB research
- Mar 2022Deutschlands Mittelstand #2: so meistern kleine und mittlere Unternehmen erfolgreich KrisenMichael Banf and Matthias Geissler · Study series of the RKW Kompetenzzentrum, Eschborn · data: Oskar-Patzelt-Stiftung, Großer Preis des Mittelstandes 2021PublicationSMB research
Contact