Fellowship – Computer Vision FEL

Advanced Technology Applications Center (ATAC), Global Databases Division, Infrastructure and Platforms Sector

Contract Duration - 12 months

Duty Station: CH-Geneva

Final date to receive applications: 08-Jun-2026, 11:59:00 PM CET

(a) Organizational Setting

Fellowships aim to provide fellows with experience to strengthen their knowledge and professional competence, which they can apply in their professional field once they have completed their fellowship with WIPO. Fellowships are awarded for specific activities, which provide the fellows training, within the approved workplan of the employing office.

This Fellowship is hosted by the Advanced Technology Applications Center (ATAC), located in the WIPO Global Databases Division, is a research and development center focused on prototyping, developing, and providing a range of Machine learning and AI powered services and tools to other database sections, to other WIPO Divisions and to other external entities. The Division has developed its own tools for WIPO and for other international organizations.

Some of the applications belong to document vision: image classification and similarity for trademarks (see http://(Use the "Apply for this Job" box below).), soon a similar functionality for the Global Designs Database will be deployed (see http://), the Division has recently developed a pilot OCR project tailored to patent documents using traditional Machine Learning. The ATAC section has also developed a machine translation set of tools (WIPO Translate engine, WIPO Translate widget), which can also be applied to graphs and tables inside a document.

Other applications include WIPO speech-to-text (voice recognition for WIPO conferences). ATAC wishes to continue exploring the applicability of AI to the above mentioned activities and other intellectual property related activities.

The ATAC Fellow will be responsible for improving the quality of the above-mentioned systems with a specific focus on computer vision applications applied to:

- patent documents, including pdf documents (native or scanned)

- general web HTML documents including graphs, tables and video subtitles

- trademarks and industrial designs images.

WIPO is using the six official UN languages (Arabic, Spanish, English, French, Russian and Chinese), and all its products must support multilingualism by design, in addition, some applications also support other languages such as the Japanese, Korean, German, Portuguese.

The ATAC Fellow will take responsibility for gathering and cleaning data, explore techniques to apply on this data, define a roadmap, build running prototypes and assess their usage in production in our areas of application.

The ATAC Fellow must be capable of taking ownership of already existing projects while using and improving them, while developing greenfield prototypes and applications. The Fellow is expected to be able to describe and argue the reasons for their code and documentation regardless of if it was created with or without AI assistance.

The Fellow works for ATAC under the direct supervision of the Senior Machine Learning Specialist.

The Fellow will perform the following principal duties:

(a) Test various open-source solutions available for computer vision:

• Test various OCR models taking into account our use cases: patent documents requiring specific document layout analysis. Including tables, images, chemical formulae, mathematical formulae etc.
• Text extraction and analysis capabilities with special focus on multilingual aspect (incl. Chinese/Japanese/Korean)
• Integrate image processing models into structured search solutions
• Integrate text extraction and machine translation
• Explore tools for image similarity and classification components (e.g. trademark logos, industrial designs, patents drawings...)
• Explore combined text and image processing approaches, evaluate their application in various image retrieval scenarios

(b) Study the current state-of-the-art technology, evaluate the pros and cons of various methods, define a strategy for experimenting with and assessing with…