Home / AI Service / Details
📊 KI-Service

KI-Tools

Humboldt-Universität zu Berlin

KI-Tools

Institution: Humboldt-Universität zu Berlin Category: AI Service
Website: https://ki.cms.hu-berlin.de/de

Description

The guideline provides a comprehensive introduction to the use of AI at Humboldt-Universität Berlin, with a focus on data protection-compliant, internal AI tools. It serves as a handbook for research, teaching, and administration to securely and efficiently integrate generative AI. The main benefit lies in providing open-source-based, locally running AI tools such as language models, transcription and image analysis tools, as well as a JupyterHub. Furthermore, practical examples and technical interfaces (APIs) for own development are offered to facilitate the onboarding process.

Access & Availability

Target Audience: -
Access Requirements: -
Availability: - Operating hours or availability: Available 24/7 (JupyterHub is freely accessible from outside the network; HPC@HU only accessible via VPN, but otherwise usable around the clock). - Beta/pilot status: HPC@HU is currently available as a test access; the platform "Azimuth" is in pilot phase. - Restrictions: HPC@HU is accessible only from within the HU-VPN; KI tools and APIs are usable only within the HU network or via VPN. - Usage limits: Rate limits are recommended for API usage (e.g. 10–15 RPM); simultaneous parallel requests should be avoided to ensure availability for all.

Contact & Support

Contact Person: hpc-support@hu-berlin.de
Email: hpc-support@hu-berlin.de
Support: -

Technical Details

  • Technology: ["OpenAI API", "OpenStack", "REST", "JupyterHub", "Whisper", "Vision Language Model (VLM)"] HINWEIS: WebDAV ist nicht in den Dokumenten enthalten und daher nicht validierbar.
  • Platform: - Web application: Yes, KI tools and JupyterHub are available as web applications (e.g., via https://ki-tools.hu-berlin.de and https://jupyterhub.hu-berlin.de).
  • App: Not specified.
  • API: Yes, OpenAI-compatible APIs for LLMs, vision models, coding models, embedders, and rerankers are available.
  • Self-hosted: Yes, the infrastructure (LLMs, HPC@HU, JupyterHub) is operated by HU Berlin itself (self-hosted).
  • Cloud: No, the platforms are not provided in a public cloud (e.g., AWS, Azure), but on HU-owned servers.
  • Integration in existing systems: Yes, JupyterHub is integrated into the HU's digital teaching and learning landscape (HDL3) and features an interface to HU-Moodle.
  • License: -

Features

  • Provision of privacy-compliant AI tools based on large language models (LLMs)
  • Use of open-source products for AI infrastructure
  • Access to a High Performance Computing Cluster (HPC@HU) for research
  • Provision of a JupyterHub for interactive Jupyter Notebooks in the browser
  • Use of Jupyter Notebooks for prototyping, data analysis, data visualization, and computational essays
  • Provision of an OpenAI-compatible API for integrating LLMs into own applications
  • Integration of AI tools into the HU's digital teaching and learning environment (HDL3), including interface to HU-Moodle
  • Support via a chat interface for general AI tasks
  • Function for transcription of audio and video files with automatic email delivery of results
  • Function for preparing transcriptions into structured, readable texts with subheadings

Recorded: 2026-01-11
Source: https://ki.cms.hu-berlin.de/de

Visit Website