Bosch Rexroth AG

Investigation of custom parsing methods for technical documents as a preparation step for vector database embedding; Fine-tuning and extension of LLMs for domain-specific use in industrial hydraulics; Evaluation of prompt-tuning methods for tailored results from LLMs.

There are three subject areas to choose from for your thesis: Development and training of an ML Model to parse industry-specific documents as entrypoint for an RAG pipeline; Development of custom parsing and chunking tools for industry-specific documents and their ingestion into a vector database; Training and/or fine-tuning of a custom LLM model for use-case specific usages within the scope of smart hydraulic services.

During this Master thesis, you will develop and train an ML model to parse industry-specific documents as an entry point for a Retrieval-Augmented Generation (RAG) pipeline. Your job includes the development of custom parsing and chunking tools for industry-specific documents, as well as their ingestion into a vector database. Last but not least, you will train and/or fine-tune a custom LLM model for use-case-specific applications within the scope of smart hydraulic services.

As part of the Sales Commercial Operations team, you will help us further develop our internal communication concept to increase awareness of our team across Bosch Rexroth. You will actively drive the expansion and networking within our internal sales community by independently creating blog posts, presentations, etc. In this context, you will view our internal GenAI tools as an opportunity and support their increased application within our team.

During your internship, you will design and implement digitization solutions for the sales of our products, including data extraction and transformation as well as participation in translation automation, software testing, analytics and dashboarding, including application and integration of AI tools.