Internship – Embeddings for Epidemic Intelligence

WOAH - World Organisation for Animal Health; Paris; Presencial; EUR 7800 - 8900; Jornada completa

Ventajas ofrecidas por este puesto de trabajo

Monthly stipend

WOAH HQ in Paris

Descripción de la vacante

WOAH seeks an internship in its Data Science Lab (DS Lab) in Paris to develop and evaluate a semantic-matching pipeline for epidemic intelligence. You will review multilingual embedding models, build pipelines in Python or R, and compare results against reference descriptions of disease events.

The project involves collaboration with the EPIQ team and WOAH regional partners. Ideal candidates are enrolled in a Master’s program in data science or related fields with experience in text processing,

Formación

• Currently enrolled in a Master’s degree or final year of related program.
• Experience with Python or R for text data processing and statistics.
• Interest in NLP, embeddings, and multilingual processing is desirable.

Responsabilidades

• Review embedding models for multilingual coverage and licensing.
• Build a reproducible semantic-matching pipeline in Python or R.
• Compare embeddings against disease-event references and open-source articles.
• Apply pipeline to EPIQ data and WAIS/open-source streams to assess added value.
• Document pipeline and methods using best practices and AI coding tools.
• Collaborate with the DIAD and WOAH teams and stakeholders.

Conocimientos

Python

NLP

LLMs

AI coding tools

Statistics

English

Reproducible code

Git

Educación

Master's degree in data science / related field

Herramientas

Claude Code

Codex

Git

Descripción del empleo

Paris, France

Context

The World Organisation for Animal Health is a leading intergovernmental organisation representing 183 Members worldwide. Through its activities, WOAH makes a decisive contribution to improving animal health, protecting animal welfare and strengthening Veterinary Services. The Organisation provides transparent information on world’s animal health situation, and promotes international standards, particularly in terms of the safety of trade in live animals and animal products. More information can be found on WOAH’s website.

Joining WOAH means taking part in the development of one of the leading international organisations, recognised and associated with other multilateral institutions, in the field of worldwide health. It means helping to build a global approach to health, combining animal and human health in a “One Health” approach. It means joining teams motivated by the impact of their actions, the sense of their collective commitment and their recognised professionalism in their respective fields of expertise.

WOAH’s headquarters are based in Paris. The Organisation is present on every continent through 13 Regional or Sub-regional Representations. WOAH has 250 staff members, two-thirds of whom are based at headquarters.

The mission of the Data Integration and Analytics Department (DIAD) is to enhance WOAH’s decision-making capacities by analysing data from diverse internal and external sources and providing evidence-based insights for WOAH, its Members and partners. Through advanced analytics, data integration and knowledge generation, DIAD supports data-driven decision-making across the Organisation. Within DIAD, the Epidemiology & Quantitative intelligence team (EPIQ) collects and analyses non-official information from open-source media to detect early, unconfirmed signals of animal disease events worldwide, ahead of and alongside official reporting through the World Animal Health Information System (WAHIS). This information is used for epidemic intelligence purposes, to strengthen international animal health surveillance and to support horizon scanning for emerging threats. EPIQ also works continuously to improve the methods and tools used to collect non-official information, so that relevant signals are captured more completely, more quickly and with less manual effort. In the Asia-Pacific region, this work is carried out together with the WOAH Regional Office for Asia and the Pacific (Tokyo), under the project “Strengthening early threat warning in the region”, funded by the Australian Government Department of Agriculture, Fisheries and Forestry (DAFF). Current event-based surveillance tools mostly identify relevant articles through lexical matching, such as keyword lists and Boolean queries. As a result, articles that describe an event with other words or in other languages can be missed, while off-topic mentions of disease names increase the time spent on triage. Dense text embeddings, especially multilingual ones, could match articles on meaning rather than exact wording, but their performance has seen little systematic testing in animal health surveillance.

In this context, DIAD is offering an internship opportunity within its Data Science Lab (DS Lab) and EPIQ team for a motivated student interested in natural language processing, artificial intelligence and statistics applied to animal health surveillance. The internship is funded by DAFF and is part of the Asia-Pacific early threat warning project. Based at WOAH Headquarters in Paris, the intern will develop and evaluate a proof-of-concept semantic-matching pipeline for epidemic intelligence.

Job Description

Positioning and reporting

The intern will join the Data Science Lab (DS Lab) and the EPIQ team of the Data Integration and Analytics Department (DIAD), under the authority of the Head of DIAD and the direct supervision of a Senior Veterinary Epidemiologist and the AI Scientist of DIAD, and of an Animal Health Information Officer at the WOAH Regional Office for Asia and the Pacific in Tokyo, who co-leads the project’s epidemic intelligence activities.

Job purpose

The purpose of this internship is to build a proof-of-concept semantic-matching pipeline, based on text embeddings, that could complement lexical retrieval in WOAH’s epidemic intelligence activities. The pipeline will be evaluated against the same benchmark and under the same conditions as the existing lexical approach, with sensitivity and specificity estimated against WAHIS-confirmed events and expert-annotated data.

Missions and activities

1. 1. Review of embedding models
   • Review candidate embedding models (multilingual, general-purpose and biomedical, open-source and commercial), with classic word embeddings as a lower-complexity reference.
   • Compare them on coverage of the languages monitored by the EPIQ team, fit with animal health and veterinary terminology, computational cost and latency at scale, and licensing and deployability (self-hosted or API-based).
2. 2. Implementation of a data pipeline based on various embeddings
   • Develop a reproducible semantic-matching pipeline (Python or R) using at least two or three embedding models, comparing incoming articles with a set of reference descriptions of disease events and flagging candidates based on semantic similarity.
   • Apply the pipeline to the information already collected by the EPIQ team and, where feasible, to an independent stream of open-source articles, to distinguish its added value for prioritising articles already retrieved from its ability to identify relevant articles missed by keyword-based retrieval.
   • Use AI-supported coding tools (e.g. Claude Code, Codex) following good practices, and document the pipeline and methods.
3. 3. Assessment of sensitivity and specificity of the pipelines, accounting for error in the calibrating data
   • Benchmark the detection performance of each pipeline against WAHIS-confirmed events and/or expert-annotated article sets.
   • Account for the imperfections of the reference data, which may be incomplete, delayed or subject to annotator disagreement, using latent-class (no-gold-standard) estimation methods commonly applied in veterinary diagnostic test evaluation.
   • Where feasible, build a manually annotated sample of news articles as an independent evaluation set, so that precision, recall and F1-score can be compared between approaches under the same conditions.
4. 4. Comparison with the existing lexical baseline
   • Apply the same evaluation protocol to the current lexical approach, or to a reproducible approximation of it.
   • Quantify where semantic matching improves or reduces recall and precision, and identify the languages, diseases and types of signals for which it adds the most value.
5. 5. Communication, interaction and coordination with the EPIQ team and the WOAH Regional Office for Asia and the Pacific
   • Interact regularly with the EPIQ team and the WOAH Regional Office for Asia and the Pacific (Tokyo) on pipeline design, data requirements and interpretation of results.
   • Present the results and recommendations to the team and relevant DIAD colleagues and the WOAH Regional Office for Asia and the Pacific, and contribute to the DAFF project reporting.
6. The Head of DIAD may, when necessary, assign to the incumbent any other duties, responsibilities and activities within the intern’s area of competence.

Requirements

Qualifications and Experience

Required qualifications and experience

• Education: currently enrolled in a Master’s degree (M1/M2) or final year of an engineering school in data science, statistics, computer science, natural language processing or a related quantitative field.
• Experience: academic or practical experience (projects, previous internship) in text data processing and statistical analysis with Python or R.

Additional experience (asset)

• Experience with text embeddings, semantic search or multilingual natural language processing.
• Knowledge of diagnostic test evaluation or latent-class models.
• Interest in or previous exposure to animal health, epidemiology, public health surveillance or the One Health approach.

Expected Skills

Technical skills

• Strong data science programming skills in Python.
• Knowledge of natural language processing, including text embeddings and large language models (LLMs).
• Experience with AI-supported coding tools (e.g. Claude Code, Codex).
• Sound knowledge of statistics, including the evaluation of classification performance (sensitivity, specificity, precision, recall).
• Good practices in reproducible code: version control (Git), documentation, testing.
• Excellent command of English (written and spoken); knowledge of French or Spanish is an asset.

Personal and Interpersonal skills

• Analytical mind, rigour and attention to data quality.
• Autonomy, curiosity and ability to learn new tools quickly.
• Critical approach to AI outputs and to the interpretation of statistical results.
• Ability to work in a multicultural and multidisciplinary team and to communicate with internal stakeholders.

Additional Information

Working conditions

• Type of contract: Internship.
• Salary: 715€ / month.
• Duration: six (6) months.
• Start date: as soon as possible.
• Location: WOAH Headquarters, 12 rue de Prony, 75017 Paris, France.

Eligibility and start date

This internship requires an agreement with the student’s academic institution.

The selection of the candidate and the internship start date will depend on the suitability of the candidate’s profile for the position, as well as on the operational and funding constraints associated with the internship.

The selected candidate must be able to commence the internship in person in Paris, France, no later than 18 January 2027, to ensure that the internship activities can be carried out within the applicable funding period.

General Information

WOAH places high value on a multicultural and positive work environment.

WOAH is an equal opportunity employer and welcomes all qualified candidates, irrespective of their origin, gender, opinions or beliefs.

20 November 2026.