89851: Data scientist, mathematician or similar (f/m/x) - Securing data quality for AI systems

Area of research:

Scientific / postdoctoral posts,Sonstiges,Promotion

Job description:The Institute for AI Safety and Security provides research and development services in the field of AI-related methods, processes, algorithms and execution environments. The focus is on ensuring safety and security for AI-based solutions in ambitious application classes. “Safety and security by design” is a central aspect in this context, since it directly supports future requirements of safety-critical applications that are based on AI or integrate AI-based components.
 Main Research Areas of the Institute for AI Safety and Security are: 

• development processes and methods for safety-critical AI applications
• robust and reliable approaches to safeguard AI methods and algorithms
• safety and security in the context of AI-based applications
• management and use of sensitive data
• execution environments for AI-based applications
• innovative computation methods.

The department Safety-Critical Data Infrastructures deals with data and information quality as well as the foundations for building secure, decentralized data infrastructures. The development of methods and technologies with a focus on data quality always takes place in relation to ensuring the security and robustness of AI systems. One focus is the investigation of relationships between the data quality of training and input data, with reference to the result quality and performance of AI systems.

The advertised position includes the following scientific topics:

• research and development on methods for determining, improving and securing data quality with regard to the performance and security of systems related to artificial intelligence.
• research and development on approaches to secure data with regard to quality and attacks in data rooms
• scientific research on studies and technical information related to approaches and challenges in the development and implementation of quality metrics for different types of data
• development of methods and approaches in the context of Exploratory Data Analysis (EDA), regarding the assessment of available data.
• integration into frameworks for AI systems, with the aim of making the researched methods directly usable in practice and available to a broad community.
• development of methods to relate the quality of (training, verification/inference) data to the results of AI models/systems.
• implement prototypical applications and representations of data quality according to current guidelines for demonstration at the institute and in projects within the scope of the project mandate.
• analysis of possible approaches to integrate technologies for the purpose of preserving and enhancing data quality in decentralized data and service ecosystems (cf. GAIA-X and IDSA).
• exploration of potential implementation pathways in big data ecosystems with heterogeneous data (source) structure.
• research and analysis of potential attack vectors and deception approaches with a focus on data.
• development of countermeasures for attacks and deception approaches.
• development of a taxonomy for data in terms of quality and potential for attack.
• development of an approach for rational assessment of risks and motives regarding attacks on data sets/instances in the data space.
• preparation of scientific (partial) contributions for publications
• derivation of options for action for scientific and industrial applications
• presentation of scientific results
• steering and management of third-party funded and core-funded projects
• technical assistance in the acquisition of national / European third-party funding

This research center is part of the Helmholtz Association of German Research Centers. With more than 42,000 employees and an annual budget of over € 5 billion, the Helmholtz Association is Germany's largest scientific organisation.