Applied Research Collaboration
From a Shared Research Question to Credible, Usable Results
We help universities, companies, and public-interest organizations structure applied climate and sustainability research around sound methods, real domain needs, and a clear path to datasets, benchmarks, publications, prototypes, or open tools.
Common barriers
What prevents research from being used
Universities, labs, and applied research projects
The science is strong; it is unclear who would use a tool day to day and for which decision.
A demo works in the lab. Stakeholders cannot integrate it into their workflow or trust it enough to rely on it.
Licensing, data sharing, and what should stay closed are decided late or not at all.
Grant language covers deliverables; productization, maintenance, and life after the project are thin.
Research partners are in place; nobody owns interfaces, user research, or implementation.
Methods, data, code, and findings are produced, but publication, documentation, licensing, and future maintenance are not planned together.
Research · Domain · Delivery
A practical collaboration model
Each participant has a distinct role, with methods, authorship, data rights, and review responsibilities agreed at the start
Researchers and PhD candidates
Academic contributors bring scientific methods, literature, evaluation design, interpretation, and research questions connected to their work.
Companies and domain stakeholders
Companies, NGOs, public bodies, and practitioners bring concrete questions, data where appropriate, domain labels, constraints, and user feedback.
Product and engineering
Climate+Tech translates the shared question into reproducible datasets, evaluation pipelines, interfaces, open-source components, and usable workflows.
Supervised student contribution
Working students may support suitable, well-defined tasks such as literature review, annotation preparation, implementation, testing, or documentation under qualified supervision.
How an applied research collaboration works
Step 1: Frame the shared question
Define the decision context, stakeholders, research question, existing evidence, and what each partner contributes.
Step 2: Design the evidence
Select methods, data, annotation protocol, baselines, evaluation criteria, expert review, and research ethics or data-governance requirements.
Step 3: Build and validate
Researchers, domain experts, engineers, and supervised student contributors develop and test the dataset, benchmark, pipeline, or prototype.
Step 4: Publish and put into practice
Prepare the appropriate publication, documentation, open release, stakeholder pilot, product workflow, or next funded research phase.
Collaboration support
What we bring
A clear operating structure in which research leadership, domain input, supervision, and technical delivery remain distinct
We turn a shared problem into defined research questions, roles, review points, data responsibilities, and outputs for academic and practice partners.
We help define data selection, annotation protocols, quality checks, baselines, evaluation metrics, and the distinction between training and independent test data.
Where the work supports an original contribution, we plan reproducibility, code and data release, documentation, attribution, and a publication or conference path with the research leads.
After validation, we help translate methods into open-source or open-core components, user workflows, stakeholder pilots, and maintainable software.
Research and practice
Possible collaboration outputs
The output depends on the research question, evidence, partner roles, and publication or adoption goal
Collaboration and research plan
Research questions, partner roles, data responsibilities, supervision, review points, authorship principles, and intended outputs.
Dataset and annotation protocol
A documented training or benchmark dataset with provenance, label definitions, quality checks, and appropriate data splits.
Benchmark and evaluation report
Baselines, metrics, error analysis, expert agreement, limitations, and reproducible evaluation code where appropriate.
Publication or dissemination path
A credible route to a paper, conference contribution, technical report, dataset paper, or public methodology note when the evidence supports it.
Open-source prototype
A documented implementation that makes a validated method testable and reusable without presenting research code as production software.
Stakeholder pilot or funded next phase
A plan for real-user validation, consortium delivery, further research, maintenance, or responsible productization.
Applied research case study
UsefulBench: from a company AI problem to published research
A core case study for collaboration between companies, professional domain experts, Climate+Tech, and university researchers
Score4More needed to improve text selection in its sustainability AI pipeline. Topically relevant passages were not always useful for answering an analyst's question.
Climate+Tech helped set up the collaboration with University of Zurich researchers, including Tobias Schimanski, around a clear information-retrieval question and evaluation design.
Professional sustainability analysts contributed real questions and labelled passages separately for relevance and decision usefulness, creating evidence grounded in practice.
The collaboration produced the UsefulBench dataset and working paper, with transparent experiments, limitations, and a path back into better retrieval workflows.
Latest from research and tools
View all posts »Related posts on climate research, open tools, and applied AI.
ESG & AI: Revolutionizing Real Estate Investment
Discover how AI is shaping the future of ESG in real estate investments for a sustainable and profitable tomorrow.
Climate Product Strategy & Scoping
Climate tools only matter if people use them. We help you clarify what to build, what not to build, who buys, and the smallest useful next step.
UsefulBench: Can Information Retrieval Find What Is Useful, Not Just Relevant?
A relevant passage is not necessarily useful for a decision. UsefulBench tests whether information retrieval systems can tell the difference in sustainability reports.
30 Data Quality Problems in Corporate GHG Emissions Reporting
A GreenDIA and Klimatkollen technical report separates extraction accuracy from source-report quality and proposes 30 errors and issues that human annotators and Climate NLP systems can assess.
Frequently Asked Questions
Who is this for?
University labs, PhD and postdoc teams, Horizon consortia, applied climate research projects, science-based NGOs, and public-interest AI projects.
Do you do academic research for us?
We focus on productization, tooling, adoption, and delivery. Scientific leadership usually stays with the research partners; we help make outcomes usable.
How can a company and university work together?
The company or domain organization brings a concrete problem, context, and feedback. Academic partners lead methods and scientific interpretation. We help define the operating model, datasets, software, evaluation workflow, and route into practice.
Can working students contribute?
Yes, when there is a suitable learning objective, clear supervision, and a well-defined task. Students can contribute to research engineering and documentation, but they do not replace domain experts, research leadership, or professional delivery.
Does every collaboration lead to a publication?
No. Publication depends on the novelty, evidence, permissions, and academic standards of the work. We can plan for reproducibility and a publication path, but research partners decide whether the result is publication-ready.
How is this priced?
Scope and price are agreed after a 60-minute scoping call. We do not publish fixed package prices.
Is this the same as joining a Horizon consortium?
Related but different. For an ongoing consortium role, see Consortium & Work-Package Partnering. This sprint clarifies the product path first.
What if we mainly need open-source adoption?
See the Open Infrastructure Adoption Sprint.
Have a shared research question?
Discuss the academic, domain, data, and implementation roles before choosing a format.
We usually reply within a few working days to schedule.
Discuss an applied research collaboration
Tell us about the research question, partners, data, intended users, and whether you are aiming for a dataset, benchmark, publication, prototype, or open tool.