013S-Project (Series I, II)
The 3S-Project focuses on advancing anti-doping efforts by addressing the critical challenge of sample swapping, where athletes exchange urine samples to evade detection. Leveraging machine learning algorithms, the project aims to enhance the sensitivity and accuracy of DNA identification techniques to confirm sample donors and detect irregularities. Building on an existing algorithm, 3S-II integrates a user-friendly tool for anti-doping authorities, offering a similarity score to identify inconsistencies within anonymized athlete profiles. Following a design science research methodology, the project evaluates the tool's impact on decision-making, fostering transparency, and empowering anti-doping organizations to ensure fair competition and integrity in sports.
More info →
02EPOPredict (Series I, II, III)
EPOPredict focuses on combating blood doping practices, particularly the misuse of recombinant human erythropoietin (rhEPO) to enhance athletic performance. This project aims to develop an indirect detection method using statistical analysis and machine learning algorithms to flag EPO abuse in blood samples. By analyzing hematological profiles and identifying key indicators, EPOPredict will establish a robust pipeline for detecting rhEPO usage. The project's goals include conducting clinical experiments, deploying state-of-the-art algorithms, and creating a comprehensive tool for anti-doping agencies. EPOPredict aspires to provide cost-effective, scalable solutions, supporting WADA's efforts to uphold integrity and fairness in sports.
More info →
03MARVIN
The MARVIN project advances anti-doping efforts by integrating advanced NLP techniques and large language models (LLMs) to enhance decision-making in anti-doping investigations. Building on insights from the 3S Project, MARVIN focuses on leveraging techniques like Retrieval-Augmented Generation (RAG), LoRA, PEFT, and fine-tuning to analyze complex data within the Athlete Biological Passport (ABP) steroidal module. By combining pattern recognition with lightweight LLM quantization methods, the project develops scalable, explainable tools for detecting anomalies in steroid profiles, such as identical patterns indicative of sample swapping. MARVIN aims to deliver cutting-edge, AI-driven solutions that empower anti-doping authorities with precision tools to uphold fairness and integrity in sports.
More info →
04HEALTH-AI
The HEALTH-AI project leverages artificial intelligence to drive innovation in genomic studies, aiming to unlock insights into complex biological systems and improve healthcare outcomes. By applying cutting-edge AI techniques to analyze vast genomic datasets, the project focuses on identifying patterns, biomarkers, and genetic variations linked to diseases. HEALTH-AI develops advanced algorithms to enhance the accuracy and efficiency of genomic analysis, enabling personalized medicine and targeted treatments. Through interdisciplinary collaboration, the project aims to bridge the gap between AI and genomics, fostering breakthroughs in understanding human genetics and supporting the development of precision healthcare solutions.
05QUASIM
The QUASIM project explores the potential of quantum computing (QC) to revolutionize manufacturing simulations, addressing challenges faced by the German manufacturing sector in maintaining global competitiveness. By integrating QC with traditional simulation methods like the finite element method, QUASIM aims to accelerate calculations and reduce the complexity of modeling efforts. Quantum Machine Learning will be employed to streamline simulations, making them more accessible, especially for SMEs with limited expertise. The project will develop innovative QC solutions, integrate them into low-threshold services, and provide access via GAIA-X environments, enabling a broader adoption of cutting-edge simulation technologies in manufacturing.
More info →
