Einstein Foundation Early Career Award: 2025 Finalists
From a pool of over 70 applicants, five pioneering research projects were nominated to present their work during the Berlin Science Week 2025. After a rigorous evaluation, the award jury selected ERRING RIGOROUSLY as the winner.
We are delighted to introduce you to the other finalists:
AI’s Impact on Scientific Cognition & Research Integrity
Understanding and Mitigating the Impact of AI on Scientific Cognition and Research Integrity, led by Laura Globig (New York University, USA)
Generative AI is rapidly transforming scientific research, but its influence on researchers’ thinking and integrity is largely unexplored. Laura Globig’s project investigates how scientists actually use tools like ChatGPT and whether AI can reinforce or undermine rigorous scientific practices. By combining large-scale surveys of AI usage across disciplines with behavioral experiments, the project will develop a fine-tuned Research Integrity Assistant.
This AI tool will provide context-sensitive guidance on open science practices, transparent reporting, and critical evaluation strategies, encouraging reflection and reducing bias without being intrusive. Through a randomized controlled trial, the assistant’s effectiveness will be tested against standard AI tools, with the aim of embedding rigorous thinking and reproducibility into daily research workflows. The project addresses an urgent metascientific challenge: ensuring that AI serves as a catalyst for improved scientific quality rather than a conduit for bias or error. Its outputs – an evidence-based map of AI use and an opensource intervention tool – will support researchers worldwide in adopting robust and transparent practices, advancing research integrity across disciplines.
→ Watch the project presentation here
Democratizing AI with FAIR data and model sharing
Image2Model Exchange: Democratizing AI analysis through FAIR data and model sharing, led by Hernan Andres Morales Navarrete (Universidad De Las Américas, Ecuador)
Image2Model Exchange is an open-source global initiative that allows researchers to exchange embryonic imaging data for AI-generated analytical models. By leveraging brightfield, widefield, and confocal microscopy datasets, the project creates a reciprocal bartering system where contributors receive ready-to-use AI tools tailored to their data while adhering to FAIR principles (Findable, Accessible, Interoperable, Reusable).
The platform democratizes access to advanced computational methods, enabling laboratories worldwide – including those with limited resources – to analyze complex developmental biology datasets and gain novel insights. Using supervised and unsupervised AI models, as well as transfer learning across species and imaging modalities, Image2Model Exchange accelerates discovery, standardizes analysis, and fosters reproducibility. Community-driven governance, feedback loops, and contributor incentives ensure sustained participation, while fully open source and open data policies maximize transparency and usability. Led by Hernan Andres Morales Navarrete, the project aims to build toward a foundational, cross-species AI model of embryogenesis, creating an inclusive, collaborative, and high-impact scientific ecosystem.
→ Watch the project presentation here
High Resolution Climate Dataset for West Africa
High Resolution Climate Dataset for West Africa: A Service for Climate Protection and Energy Transitions, led by Aissatou Ndiaye (University of Augsburg, Germany)
This project aims to deliver the first 3 km-resolution climate dataset for West Africa, providing the detailed information needed for climate protection and energy transition planning. By downscaling CMIP6 climate models with the WRF-Solar model, the project produces hourly data on temperature, wind, solar radiation, and other key variables for both historical and future climate scenarios.
The dataset supports renewable energy projections, informs policymakers, and strengthens regional climate resilience planning. Open-source modeling frameworks and GitHub repositories ensure reproducibility, transparency, and collaboration, while regional partnerships enhance local ownership and capacity-building. Led by Aissatou Ndiaye and a geographically diverse team from Germany, Rwanda and Senegal, the project addresses a critical knowledge gap in a climate-vulnerable region, enabling data-driven decision-making and supporting equitable, sustainable energy transitions.
Team Members: Victoire Ghafi Kondi Akara (African Institute for Mathematical Sciences Research and Innovation Centre, Rwanda), Windmanagda Sawadogo (Karlsruhe Institute of Technology, Germany), Oluwarotimi Delano Thierry Odou (International Renewable Energy Agency), Dahirou Wane (University Cheikh Anta Diop of Dakar, Senegal)
→ Watch the project presentation here
Open Data & Reproducibility in MSK Ultrasound
UMUD: A Web Application to Address the Need for Open Research Data and Reproducibility in Musculoskeletal Ultrasonography, led by Paul Ritsche (University of Basel, Switzerland) & Fabio Sarto (University of Padova, Italy)
UMUD is a centralized repository for musculoskeletal ultrasound datasets, addressing critical gaps in data standardization, accessibility, and reproducibility. The platform aggregates metadata, benchmark datasets, and standardized pipelines for automated image analysis, fostering transparency and enabling AI-based research in muscle, tendon, and ligament assessment.
By providing FAIR-compliant, open-access resources, UMUD promotes reproducible research and supports clinical translation. Planned activities include expanding datasets and benchmarks, establishing standardized image acquisition protocols, and engaging the research community. Led by Fabio Sarto and Paul Ritsche, UMUD empowers researchers to collect high quality data, enhances open science practices in musculoskeletal research, and serves as a model for other imaging domains.
→ Watch the project presentation here