collaborate

Collaboration

This page is the entry point for collaboration with the research group of Prof. Dr.-Ing. Michael Färber (ScaDS.AI/TU Dresden).

Our core topics include trustworthy AI, LLMs / chatbots / question answering, retrieval-augmented generation (RAG) and evidence-based generation, knowledge graphs, and graph machine learning. We are also open to side topics that connect to these areas, such as agentic and multimodal AI, robustness & security (e.g., prompt injection in RAG-style systems), causal reasoning, AI for science, and quantum natural language processing.

Collaboration formats (typical)

• Publicly funded projects (preferred when possible): joint proposals (e.g., DFG/BMFTR/EU calls), with clear milestones and publishable outcomes.
• Industry / public-sector pilots (4–12 weeks): scoped feasibility studies, benchmarking, or prototypes that de-risk a larger project.
• Contract research / applied prototypes (3–12 months): implementation and evaluation under agreed constraints (incl. NDA/IP).
• Student theses / student projects (co-defined topics): strong fit when the topic is well-scoped and results can be documented (often publishable).
• Research stays / visiting researchers: short visits to collaborate on a concrete research question.

Good first step: a small pilot

We often start with a small, well-scoped pilot to align expectations and “test the waters” before moving to larger funding or long-term collaborations.

Typical pilot deliverables (examples):

• a short technical report (findings + recommendation)
• a benchmarking / evaluation of candidate methods
• a prototype (reproducible code + minimal demo)
• a project plan for a larger funded proposal

Publications, IP, and data

• Publication-first by default: whenever feasible, we aim for publishable results and open artifacts (code/data) in line with academic best practice.
• Confidentiality/IP: NDAs and review periods are possible; we clarify constraints early so projects remain feasible.
• Student theses: as a rule, theses should be publishable; delays can be agreed for IP/patent or sensitive data, but fully non-publishable thesis topics are typically not a good fit.
• Data access: please state early whether data is sensitive (personal data, trade secrets, security constraints) and how access can be provided (on-site, secure VM, anonymized subset, etc.).

If a collaboration is publicly announced, this should be coordinated in advance.

What to send (1-page brief)

Please send a short email including a 1-page brief with:

• problem statement and context (what is the pain point?)
• what success looks like (deliverables)
• available data + constraints (privacy, access, IP)
• desired timeline and budget range (rough is fine)
• who will be involved on your side (roles)

Contact

For contact details, please see /contact/.
For press inquiries, see /media/ and /press-kit/.
Selected funding & partners: /funding/.