LLMs at NHR@FAU

Title: LLMs as a Service at NHR@FAU

About the speaker: Armine Garibyan (FAU Erlangen-Nürnberg) has experience in rule-based gesture detection within large video data repositories, utilizing the Woody cluster (CPU) for computational efficiency. She has also trained LoRa adapters and conducted full fine-tuning of large language models (LLMs) for linguistics tasks using the Alex cluster (GPU). Her main areas of expertise include video data analysis, LLM development, and the application of machine learning in computational linguistics.

Abstract

Large Language Models (LLMs) have become an integral part of many researchers’ daily workflows. While occasional use through a chat interface is often sufficient for simple tasks, research projects frequently require hundreds or even thousands of LLM queries. In such cases, API access is essential for automating experiments, processing larger datasets, and integrating LLMs into research pipelines. Commercial LLM APIs often impose rate limits or usage-based fees, making large-scale use costly and, in some cases, impractical. To address this challenge, NHR@FAU provides LLM-as-a-Service, offering academic users access to state-of-the-art language models without such restrictions. This session introduces the LLM endpoint available at FAU and demonstrates how it can be used in research projects. Practical examples from the humanities and social sciences will illustrate common use cases, and live demonstrations in both R and Python will show how to interact with the API and integrate LLMs into existing workflows. No prior experience with LLMs or APIs is required. The session will conclude with a hands-on demonstration during which participants will have the opportunity to make their own API requests and explore the service themselves.