Large Language Models (LLMs), e.g., ChatGPT, show impressive performance in building state-of-the-art applications such as conversational AI assistants. However, LLMs fail unpredictably in knowledge-intensive tasks, i.e., in scenarios where factually correct answers must be guaranteed. Knowledge Graphs (KGs), on the other hand, can represent knowledge as triples and have complementary strengths to LLMs’ weaknesses. As a member of the Data Science and Artificial Intelligence department, you will support us in designing empirically well-grounded methods for combining LLMs and KGs in both directions, i.e., using LLMs to automate the construction of KGs, and the use of KGs to improve the unsound answers that LLMs often produce. Your employment should start as soon as possible.
| Job Type | HiWi |
| Extent | 15-19h |
| Status | Open |
| Contact(s) |
Your tasks include:
- Design and implement approaches and tools for using knowledge graphs (KGs) in combination with large language models (LLMs) for different applications including conversational AI and content generation.
- Development of agile data integration methods with KGs to link data sources to become the source of knowledge for LLMs.
- Development of new methods to improve training data and evaluate data quality for LLMs based on KGs.
- Analysis of user requirements as well as support and professional consulting for our customers from the business community.
- Software development in Python
What you bring to the job:
You have completed your university studies (bachelor’s degree) in Computer Science or a neighboring field and bring a passion for application-oriented science to the job. In addition, sound knowledge in one or more of the following areas is required:
- Knowledge of machine learning models (especially deep learning architectures). Experience with transformer architectures and libraries from Hugging Face is a plus.
- Knowledge in the areas of semantic technologies and knowledge graphs.
- Experience in data management for machine learning, e.g., data preparation, and quality analysis of training data.
- Experience in natural language processing methods and libraries
- Excellent programming skills in Python
- Experience with software development and deployment tools and practices, i.e., GitHub or GitLab, clean code practices, REST APIs, and Dockers
- Experience with teamwork, ideally also in changing teams and deadline-driven projects
- Very good knowledge of German and English, both written and spoken
What we offer:
- We focus on innovative ideas for practical applications. You can be part of our mission and actively shape the future.
- You work in a leading international institute for applied research with us.
- You can explore and use state-of-the-art technologies and methods.
- If you are interested in pursuing a Ph.D. within our research areas, you can get support.
- Your personal development is important to us: we support you and offer helpful learning.
- Flexible working hours tailored to your studies (max. 19 hours/week).
Severely handicapped persons will be given preference in the case of equal aptitude. Fraunhofer-Gesellschaft attaches great importance to gender-neutral professional equality.
Interested? Then send your résumé (English or German) to:
Dr. Diego Collarana
Group Leader Knowledge-Enhanced Large Language Models
Email: diego.collarana.vargas@fit.fraunhofer.de
Tel.:+49 2241 14-3617