Research Assistant / PhD Candidate for Natural Language Processing (NLP) and Causal AI (m/f/d)

The Technical Information Library (TIB) is a foundation of public law of the State of Lower Saxony. With around 600 employees and an budget of around 60 million euros, it is one of the largest information infrastructure facilities in Germany. As a German Central Library of Technology and Natural Sciences, our pioneering services ensure the infrastructural requirements of high-quality information and literature supply for research in science and industry. With the Open Research Knowledge Graph (ORKG), we work to revolutionize the exchange and use of scientific knowledge in the digital age. The TIB, Programme D, Open Research Knowledge Graph, is looking for the Data Science & Digital Libraries (Prof. Dr. Sören Auer) research group at the next possible time:n Research assistant:in / PhD student:in für Natural Language Processing (NLP) und kausale KI (m/w/d) The job is limited for the duration of the project period of three years. The adjustment is made subject to the final approval of the funds by the third-party donor Leibniz Association. The regular weekly working time is 39.80 hours (full time). The job is generally suitable for parttime employment. Mobile work is possible to a limited extent, taking into account service requirements. The grouping takes place in the remuneration group 13 TV-L.

The site is part of the project EcoAIRE - AI-assisted hypothesis assessment in invasion biology and the restoration of freshwater ecosystems under fragmented and heterogeneous ecological evidence in the Leibniz cooperative excellence programme. The project combines the expertise of TIB in the processing of natural language and the scientific knowledge representation with the ecological expertise of the Leibniz Institute for Water Ecology and Inland Fisheries (IGB) and the expertise of the German Research Center for Artificial Intelligence (DFKI) in the causal modelling. At TIB, your research focus is on the processing of natural language for the extraction of scientific and causal information. You develop methods that translate scientific evidence into structured, machine-processable representations of hypotheses, variables, context information, uncertainties and causal relationships. Your research goes beyond the extraction of individual facts: you model how scientific findings are related and under what conditions scientific statements apply. Building on this, you will explore causal approaches inspired by world models for scientific leaders. These include simulations, intervention-based conclusion procedures and the assessment of hypotheses under changed conditions. Your activity includes You develop a multimodal framework that integrates heterogeneous scientific sources of knowledge - including publications, images, supplementary information, data sets and other research artefacts - into structured representations for hypothesis evaluation, taking into account the origin of the information. You develop language model-based and agent-based methods to extract hypotheses, variables, context information, evidence, uncertainties as well as causal or mechanistic relationships across sources, to unite, to match and to merge. You explore causal approaches, inspired by world models, for scientific followers to simulate and investigate how far hypotheses can be transferred to changed conditions. You develop evaluation frameworks that cover different quality aspects of AI-based hypothesis assessment. These include procedures based on expert assessments and approaches in which large language models are used as an assessment entity (LLM-as-a-judge). You publish your research results at conferences and in journals with peer review procedures in the fields of NLP, Semantic Web, artificial intelligence and machine learning. You will also participate in project reports and the dissemination of research results.

Complete academic studies (master or equivalent degree) in computer science or a closely related field, for example artificial intelligence (artificial intelligence), mechanical language processing (natural language processing), machine learning (machine learning), data science or computer linguistics. Detectable experience in research or software development with Python as well as practical knowledge of modern methods of software development, such as git-based collaboration, automated testing and reproducible experiments. Detectable practical experience with transformer-based models and major language models, including the use of frameworks such as PyTorch, Hugging Face Transformers or comparable technologies for NLP applications. Practical experience in the development of agent-based AI workflows or multi-agent AI workflows, including tool usage, structured expenditure, orchestration or iterative processing pipelines for conclusions. Very good knowledge of English at least level C1 for scientific communication, cooperation and publication. Required qualifications Experience with fine tuning or adaptation of large language models and/or the establishment of advanced RAG pipelines using technologies such as PyTorch, Hugging Face Transformers, Sentence Transformers, LangChain, LlamaIndex, FAISS, Qdrant or comparable tools. These include embedding-based information search, reranking, vector search and structured generation. knowledge of causal modelling and causal inference, such as causal graphs, structural causal models or intervention-based conclusion procedures. Experience with Semantic Web technologies. Expertise in ecology, in particular freshwater ecology, invasion biology or renaturation ecology, or experience in cooperation with interdisciplinary scientific projects.

Our aim is to rethink and innovate the provision and use of research data and information. We offer an intellectually inspiring environment with entrepreneurial thinking, embedded in a leading technical university and one of the largest information centres in the Leibniz community. With the L3S research centre at Leibniz Universität Hannover, one of the world's leading research institutes in the field of Web & Data Science, there is a close cooperation within the framework of the Leibniz Joint Lab Data Science & Open Knowledge. Last but not least, we value an open and creative working environment where it is fun to work. In addition, we offer a community-oriented workplace in the public service on the basis of the TV-L with a remuneration according to the remuneration group 13 TV-L. a special payment at the end of the year and 30 days holiday a year at a five-day week as well as additional free days on Christmas Eve (24.12.) and New Year's Eve (31.12.). a flexible workplace in time and space with offers for the reconciliation of work and family, such as mobile work and flexible working time models. a modern workplace in a central location of Hanover with a collegial, attractive and versatile work environment. an employer with a wide range of continuing and continuing training, occupational health promotion and pension provision for the public service (VBL). Employee discount in the students of the Studentenwerk Hannover as well as a favourable opportunity to use the versatile offers of the Hannover University Sports. independent and forward-looking activities that offer variety and leave room for personal development. the financing of the necessary equipment and of travel to conferences and research stays. cooperation in a national research and innovation project. a wide range of research infrastructures and technology components on which you can build. This includes TIB AIssistant, the Open Research Knowledge Graph (ORKG), ORKG Ask and research software from the SciKnowOrg ecosystem. .