ETEL-AI leverages cutting-edge, multi-agent AI pipelines to deliver equitable, multilingual, and quality-assured training support for UN capacity-building programmes globally. Co-designed by the Swiss Federal Institute of Technology Zurich (ETH Zurich), the Swiss Federal Institute of Technology in Lausanne (EPFL) and the United Nations Institute for Training and Research (UNITAR), this initiative bridges the gap between critical local educational needs and advanced technology.
- While UNITAR trains thousands of global professionals annually, developing curriculum-aligned materials and providing personalized feedback remains a highly manual process, often restricted by language availability and equity constraints.
- ETEL-AI adapts "Ethel"—an open-source AI feedback platform already proven across five semesters at ETH Zurich and EPFL—into a specialized "living lab" with UNITAR.
- The platform uses open multilingual LLMs to translate materials into underrepresented languages and automatically checks all content against strict psychometric standards for fairness and difficulty.
- It equips global instructors with tools for scalable personalized feedback and serves as a reusable, system-wide open-source platform focusing on entrepreneurship, digital transformation, and sustainable development.
This programme is being developed in partnership with ETH Zurich, EPFL and the UN International Computing Centre (UNICC), with total funding of CHF 349,020 requested from the International Geneva AI Innovation Program (IGAIP).
Programme Snapshot
Key facts as of July 2026:
- Duration: ETEL-AI is an 18-month project.
- Status: Application Accepted to IGAIP on 22nd June 2026 (CHF 349,020 requested);
- Planned reach: an estimated 2,000 youth and women professionals across Anglophone sub-Saharan Africa, through UNITAR’s forthcoming digital skills and entrepreneurship programme.
- Partners: ETH Zurich (PI Dr. Gerd Kortemeyer), EPFL, UNITAR, and the UNICC.
- Foundation: builds on Ethel, an open-source AI feedback platform already deployed across five semesters at ETH Zurich and EPFL.
- Adapt the Ethel AI-assisted learning platform to UN training contexts through iterative co-design with UNITAR instructors and learners.
- Generate training content through a multi-agent pipeline validated against professional psychometric standards for item difficulty and fairness.
- Extend AI-assisted content generation and learner support to underrepresented languages, using open multilingual LLMs.
Programme Phases
Phase 1 — Prototype and co-design (Months 1–8). Ethel is adapted for a first UNITAR training programme; the content generation pipeline and psychometric evaluation layer are developed and validated for one domain and language through living lab workshops with UNITAR.
Phase 2 — Expansion and integration (Months 8–14). The pipeline is extended to additional domains and languages, fine-tuned multilingual models are integrated, and the psychometric calibration layer is tested with pilot data.
Phase 3 — Deployment and handover (Months 14–18). The toolkit is packaged for independent deployment, piloted with UNITAR, and released as open source under Apache 2.0 for reuse across the UN system.
- 5 semesters of prior deployment: the Ethel platform has already been tested and refined in real classrooms at ETH Zurich and EPFL.
- 4 institutions, one living lab: ETH Zurich, EPFL, UNITAR and UNICC co-design the platform together from the outset, rather than adapting a finished product after the fact.
- Open source, system-wide: all code and model weights will be released under Apache 2.0, so any UN agency with a training mandate can reuse them.
Contact Us
For more information, please reach out to Jean Louis Raghid Kreidy at jeanlouis.KREIDY@unitar.org or Michael Adalla at Michael.Adalla@unitar.org.