AI R&D Landscape Project
Building a modern data-driven decision-support infrastructure for AI research and development
Building a modern data-driven decision-support infrastructure for AI research and development
The DACH Region and Europe
- About usWe are an international think-and-do tank with 15+ years of experience in metascience, research management, science communication, and international higher education.
Our team members have a successful track record of planning and implementing comprehensive institutional development and strategy projects across Europe, the Middle East, and Southeast Asia, including providing consultancy to national ministries and governments in three countries. - About the projectThe project goal is to create a data-driven map of artificial intelligence research and development (R&D) across the DACH region — identifying leading organisations and expertise areas, analysing talent and collaboration flows, and highlighting opportunities to establish high-impact R&D partnerships.
Why Work with Us
Why Companies Work with Us
- Clear focus on AI R&D ecosystems: research groups, collaboration networks and talent pools
- Proven ability to translate complex research data into strategic decisions
- Deep understanding of how organizations identify partners, topics and innovation opportunities
- Experience working across academia, industry and public secto
Track record (selected):
- 30+ expert and research group searches delivered
- 25+ AI and innovation domains mapped and benchmarked
- Strategic analytics projects for leading universities, corporates and public institutions
Trusted by:
- Top-500 universities
- Companies with $10B+ revenue
- Public sector organizations in 5 countries
How stakeholders can benefit from the AI R&D Landscape data
(benifit #1)
For corporate leadership, university administration and policymakers
(01)
Targeted identification of research partners
Use data to find the world’s leading research groups and institutions in your specific field, rather than relying on personal networks or reputation
Use data to find the world’s leading research groups and institutions in your specific field, rather than relying on personal networks or reputation
(02)
Understand and leverage talent flows
Track where AI talent is migrating, where talent pools are forming, and where brain drain is occurring—serving as a basis for recruitment strategies and location decisions
Track where AI talent is migrating, where talent pools are forming, and where brain drain is occurring—serving as a basis for recruitment strategies and location decisions
(03)
Assess your competitive position
Objectively benchmark your organization, region, or country against national and international competitors in terms of research output, research transfer, and collaboration density
Objectively benchmark your organization, region, or country against national and international competitors in terms of research output, research transfer, and collaboration density
(04)
Map collaboration networks
Understand who is collaborating with whom, where strong clusters are emerging, and where bridges between academia and industry are missing
Understand who is collaborating with whom, where strong clusters are emerging, and where bridges between academia and industry are missing
(05)
Continuous support within a structured environment to maintain motivation, track achievements, and refine individual learning strategies.
(06)
Spot thematic trends early
Use keyword and topic analysis to see which research fields are gaining (or losing) momentum before these trends become widely apparent
Use keyword and topic analysis to see which research fields are gaining (or losing) momentum before these trends become widely apparent
(benifit #2)
For AI experts and researchers
(01)
Find potential collaboration partners
Analyze collaboration networks to identify researchers working on complementary topics
Analyze collaboration networks to identify researchers working on complementary topics
Why analyse AI research?
- The unique field of science that rapidly transforms all other disciplines and humanity in general
- Start of the pathway to AI implementation (papers→models, libraries, datasets, repositories)
- Top venues are targeted by top businesses around the world, not only universities
- Papers at these venues are symbolic capital signals at the center of global competition for top talent and AI research evaluation
A unique, rapidly developing field of research in which the business sector is represented just as actively as universities and research institutes
New framework to measure AI research
- -1-AI research moves too fast and
is not properly captured by
paper analytics databases like
WoS or Scopus or OpenAlex - -2-It is extremely important due
to being transformative, so
stats on who’s doing the top
AI research are vital signals for
the broad spectrum of actors
(businesses, universities,
governments, society) - -3-
Lots of vibe-coded
attempts to count “top
conference papers” lead to
errors and confusion (like
the recent “tiny EU share at
ICLR”)
How do we analyse AI research?
(01)
Focus on top level AI conferences and journals
How do we define the top level AI venue?
We do not develop our own classification of leading AI conferences and journals; instead, we examine which top level AI conferences and journals are most frequently cited in job postings from tech companies.
For example, data coming from four public applicant-tracking APIs (Greenhouse, Ashby, Lever, Workable) for 90 AI companies, 7.694 postings first published in 2025 or 2026, shows that NeurIPS, ICML, and ICLR are the conferences most frequently mentioned in job postings from tech companies.
(02)
Traditional data infrastructure
The existing traditional infrastructure supporting data-driven decision-making in research and development took shape long before the AI boom and does not align with the realities of the rapidly evolving AI landscape.
For instance, organizational profiles and their sets were created for traditional scientific fields, but not for AI research and development.
Our modern data infrastructure
- More up-to-date data: indexing immediately following major AI conferences
- More comprehensive database of organizations: AI startups, labs, etc.
- Higher accuracy of entity linking and data completeness
What makes our approach relevant
Most of the analytical tools currently available for analyzing AI research:
- Aggregate large volumes of unfiltered publication data
- Treat all papers as equal, without quality control
- Offer limited filtering by country, industry or AI domain
- Are based on either journals or conferences (not both)
- Do not capture collaboration networks
- Rank people instead of research output, leading to distorted results when AI researchers change institutions
Our approach:
- Curated datasets with relevance filtering from the start
- Focus on high-quality AI research only (curated venues and sources)
- Granular filters by country, industry and AI domain
- Multiple signals integration: AI conferences, journals, GitHub, HuggingFace, etc.
- Collaboration and co-authorship networks analysis
- Research evaluation based on publications and contributions, not just individual profiles
What does this achieve?
A higher signal-to-noise ratio for strategic AI development decisions
A higher signal-to-noise ratio for strategic AI development decisions
MAPPING GLOBAL AI TALENT POOL
* Researchers who have published in leading AI venues since 2018 and whose most frequent publication venue is ICLR, ICML, or NeurIPS. Authors that only have papers in extra large teams (>30 authors) are omitted from the average team size calculation
- We identified >230.000 distinct AI researcher profiles*, representing substantial growth (+21%) from ~190.000 in 2024
- Countries vary by average researcher productivity vs team size
- We assign a country to a person using their most recent paper, so those attracting top talent (e.g. Singapore) are high by average papers per person, and those losing it are lower
- DACH countries are in the middle for both metrics
- AI talent migration data shows that between 2018 and 2024 ~270 AI researchers who were affiliated with Germany moved to the US, 130 to the UK, 100 to Switzerland and 70 to China.
- Note the UAE MBZUAI rush towards top AI talent hire, sharply focused on ICML/ICLR/NeurIPS authors
* Researchers who have published in leading AI venues since 2018 and whose most frequent publication venue is ICLR, ICML, or NeurIPS. Authors that only have papers in extra large teams (>30 authors) are omitted from the average team size calculation
How the data can help shaping the AI R&D policy?
The narrative, widely popular among policymakers in Germany and many other European countries, that Europe faces no issues with cutting-edge research but struggles with translating results into practice, attracting capital, and so forth, becomes less convincing when applied to AI research: the lag behind competitors from other regions of the world can no longer be ignored
The narrative, widely popular among policymakers in Germany and many other European countries, that Europe faces no issues with cutting-edge research but struggles with translating results into practice, attracting capital, and so forth, becomes less convincing when applied to AI research: the lag behind competitors from other regions of the world can no longer be ignored
CONNECTING AI RESEARCH TO THE REAL WORLD
AI papers matter, but models do the real work
That’s why we capture and analyse Hugging Face downloads
That’s why we capture and analyse Hugging Face downloads
- Democratic but concentrated area: top 1.000 model suppliers (out of millions) account for >95% of downloads
- Downloads can be considered as a valuable proxy for measuring impact
- We link the downloads to real-world organizations and individuals and match them to countries
- We analyse the intended usage types for models, which enables us to identify the leaders in each field and the most rapidly developing use cases
Connection between top-level AI R&D and the real world
A case of the rapid growth of the sentence transformers ecosystem of models and tools
- Developed by Iryna Gurevich, the most distinguished German NLP researcher, and her team at TU Darmstadt
- Crucial for NLP (semantic search etc.)
- Currently maintained by Hugging Face company itself
- Account for >80% of Germany’s monthly model downloads
- Other German leaders by model downloads are companies, with universities trailing behind
How is cutting-edge research translated into the real world at a personal level?
Notably, one of Iryna Gurevich’s team leaders, Nils Reimers, became VP of AI Search at Cohere
Notably, one of Iryna Gurevich’s team leaders, Nils Reimers, became VP of AI Search at Cohere
CONTACT US
Sign up here to access more data, to receive information about the report announcement or to discuss a potential partnership with us
You agree to our Terms and Conditions
