Treves, Rhineland-Palatinate, Germany
▪ Spearheaded the usage of LLMs in the company through on of the first major internal AI tools ▪ Contributed to a Retrieval Augmented Generation (RAG) system with over 10,000 anticipated international users ▪ Translated business requirements into technical specifications ▪ Collaborated with stakeholders to develop a targeted evaluation dataset to ensure the Retrieval Augmented Generation (RAG) system meets real-world business requirements ▪ Conducted in-depth analysis of the data domain ▪ Implemented query processing components and a re-ranking module in the RAG pipeline, significantly improving the retrieval accuracy and recall ▪ Implemented an evaluation framework and the visualization of the systems performance
▪ Part of a major company-wide initiative for introducing the use of LLMs to external stakeholders by gathering use cases and presenting the applicability of AI-based tools ▪ Engaged stakeholders in discussions to understand process needs and evaluate AI implementation opportunities, collaborating with internal stakeholders to analyze existing business processes and determine the applicability of LLMs ▪ Developed custom prompts for business use cases, successfully translating complex requirements into efficient AI-assisted solutions ▪ Conducted stakeholder presentations on AI implementation results, delivering insights on process improvements and potential scaling opportunities
TWin of Online social Networks (twon-project.eu): ▪ Engineered and implemented the backend infrastructure for the API by employing FastAPI and MongoDB database servers ▪ Designed and optimized prompts for LLM (Large Language Model) agents to simulate realistic user behavior in Online Social Networks (OSNs), enhancing the authenticity of simulations Social Sentiment in times of Energy Crisis (socialsentiment.org): ▪ Improved the results with Domain Adaptation by leveraging the transformers library for fine-tuning models ▪ Played an active role in strategizing and refining the data collection process, ensuring high-quality datasets to support sentiment analysis research ▪ Conducted data annotation and labeling for a large-scale Twitter dataset ▪ Gained hands-on experience in data preprocessing, annotation guidelines, and quality assurance processes ▪ Designed and maintained robust server environments, ensuring high availability and performance for institute infrastructure ▪ Deployed and managed database servers and virtualized environments using Proxmox and Docker, optimizing resource utilization and scalability ▪ Configured and managed nginx to deliver reliable and secure web services, enhancing user experience and system efficiency
▪ Overseeing the internet presence of the Central Student Advisory Office Heidelberg with the web content management systems Drupal and Imperia ▪ Design of flyers and posters with InDesign & Photoshop ▪ Assistance in Student Counceling