# ontolux

> ontolux is a reviewed AI agency in Berlin, Germany, listed in the AIAgencies.eu register and last assessed 2026-09-24.

Canonical page: https://www.aiagencies.eu/agency/ontolux
Generated: 2026-09-28 · Curated by Gabor Kiss · No paid placements — inclusion, score, and rank are never sold.

## Facts
- Location: Berlin, Germany
- Website: https://www.ontolux.de
- Hourly rate: not published
- Team size: 10-49
- Services: AI Development, AI Consulting, Generative AI
- Industries: Public Sector, SaaS & B2B, E-commerce
- LinkedIn: https://www.linkedin.com/company/neofonie

## About
ontolux is a Berlin NLP and semantic-search team within Neofonie, building German-language text analysis, tagging and search systems for public-sector, media and enterprise knowledge collections.

## Documented clients
Documented engagements: PMG Presse-Monitor, ZDF, Humboldt-Universität zu Berlin (H-Soz-Kult), BSI (Federal Office for Information Security), DEval (German Institute for Development Evaluation), Gruner + Jahr Digital Products, HUK-COBURG, Bechtle, Creditreform, Thüringer Landesrechenzentrum, and Bezirksamt Berlin-Mitte. All 11 carry a source URL on file.

## Curator assessment
Reviewed 2026-09-24 by Gabor Kiss against published evidence.

ontolux's case pages name 13 clients, and the text-analysis work carries the production weight: its TXTWerk framework tags the PMG press database (170,000+ articles a day, client since 2012) and suggests keywords and linked entities inside the ZDFmediathek's CMS, and a 2024 platform for Humboldt-Universität's H-Soz-Kult runs Llama-based semantic search on the university's own GPUs.

Three of the 13 are Elasticsearch site-search projects (HUK-COBURG, Bechtle, Creditreform), and several older ones were commissioned from parent company Neofonie, including a 2014 holiday-rental crawl for Bezirksamt Berlin-Mitte.

Few case pages report outcome numbers, and the LLM work beyond Humboldt is published as prototypes—one of which, a pharma RAG prototype, says plainly that its approach had too little impact.

The weaker fit is buyers who need a generative-AI or agent system with a production reference: the published LLM record is one local-model research platform and three prototypes.

### Key strengths
- Long client tenures stated on the case pages: Creditreform since 2009, PMG since 2012, Bechtle since 2017, ZDF and BSI since 2021
- Stack named per project—TXTWerk, Haystack, Elastic Stack, PoolParty and Doccano for the BSI threat-report system; Llama, Elasticsearch and Solr on local hardware for Humboldt-Universität
- Prototype write-ups report what did not work: the pharma comparator-therapy prototype concludes its data source had too little impact, and the client moved to a process survey before the next round

### Good to know
- Most case pages report no outcome figures—ask for usage and quality numbers on the ZDF and PMG tagging, the systems closest to production evidence
- The legal entity behind the brand is Neofonie GmbH and several projects were commissioned from Neofonie—ask whether your project is staffed from the 20-person ontolux team or the parent's 180 staff
- If you are buying LLM or RAG delivery, ask for a production reference beyond the Humboldt-Universität platform; the other LLM pages are prototypes

---
Source: AIAgencies.eu — the curated register of European AI agencies. Rate provenance: "agency-confirmed" = disclosed via submission or claimed profile; "directory-listed" = Clutch or two agreeing B2B directories (lone unverified sources are never written). Full method: https://www.aiagencies.eu/methodology · Rate index: https://www.aiagencies.eu/rates