Internship – Data Analyst multi-disease anomaly detection system using WAHIS official data and EIOS open-source information, DIAD - WOAH HQ

World Organisation for Animal Health; Paris; Presencial; EUR 10.000 - 16.000; Jornada completa

Descripción de la vacante

World Organisation for Animal Health (WOAH) invites an intern to join the epidemic intelligence team in Paris to explore anomaly detection on WAHIS and EIOS data. You will contribute to a reproducible workflow, benchmark methods, and help develop dashboards and alert products for internal WOAH use and for Members.

The internship provides exposure to time-series analysis, cross-disease benchmarking, and collaboration with senior epidemiologists, with mentoring from named WOAH staff.

Responsabilidades

• Identify and benchmark anomaly detection methods across WAHIS and EIOS data by disease and geography.
• Develop a reproducible, version-controlled pipeline to apply chosen methods and update dashboards.
• Design dashboards and automated alerts for internal WOAH use and for Members.

Descripción del empleo

The World Organisation for Animal Health (WOAH – founded as OIE) 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.

Job Description

Positioning and reporting

The Intern will report to Paolo Tizzani ([e-mail]) and Soline Chaumien-Tabouis ([e-mail])

Job purpose

Since its foundation, WOAH has been collecting information on animal diseases. As knowledge of pathogens and diseases is crucial for animal and human health, WOAH asks all its Members to report the occurrence of specific pathogens and diseases on the official list in a timely and complete manner. This information is collected and displayed on a specific public interface, the World Animal Health Information System (WAHIS). WAHIS collects official information from 183 WOAH Members and around 20 non-Members, providing a global snapshot of the animal health situation, with a continuous series of disease events available since 2005.

WOAH complements this official information with signals detected from open sources (media, official bulletins, specialised websites). This open-source screening, known as event-based surveillance, allows WOAH to detect potential animal disease events at an early stage and to verify and complement official reporting. To carry it out, WOAH takes part in the Epidemic Intelligence from Open Sources (EIOS) initiative, coordinated by the World Health Organization.

WOAH’s epidemic intelligence therefore relies on two complementary streams of information: official data reported by Members through WAHIS, which are robust and reliable information but may be subject to reporting delays and to the heterogeneity of national surveillance systems; and non-official information collected mainly through EIOS, which is timelier but unverified, not always reliable, noisier and dependent on media coverage. Detecting, as early as possible, anomalies of the epidemiological situation from its expected behaviour would allow WOAH, its Members and its partners to anticipate epidemic waves, and to target risk communication, preparedness and capacity building activities accordingly.

WOAH’s objective is to develop an anomaly detection system that exploits both streams and is applicable across all the listed diseases.

This objective is part of the framework developed by WOAH for its epidemic intelligence activities, under the activity dedicated to the detection of temporal aberrations for early warning and rapid reaction. This activity foresees the validation of an analytical approach able to warn of i) the beginning of an epidemic wave, ii) the approach of the epidemic peak and iii) the decreasing post-peak trend of the wave, using trend series from both official data from WAHIS and unofficial information from EIOS.

WOAH tested this approach on High Pathogenicity Avian Influenza (HPAI). WOAH developed, a proof of concept applying two anomaly detection analysis to the HPAI outbreaks reported in poultry to WAHIS between 2005 and 2024: the Epidemic Volatility Index (EVI) and the Convergence Epidemic Volatility Index (cEVI). This work has produced a set of assets that will be made available to the intern.

WOAH now intends to move from this proof of concept to the implementation of a, multi-disease anomaly detection tool, exploiting both WAHIS official data and EIOS open-source information, for all WOAH listed diseases.

The overall mission of the intern will therefore be to identify, benchmark and evaluate anomaly detection methods applicable to trends derived from WAHIS official data and to EIOS open-source information for all WOAH listed diseases, to combine the signals derived from both streams, to develop the resulting early warning workflow, and to make its outputs accessible in a specific tool for WOAH and, eventually, its Members.

The objectives of this project are:

• To build, from WAHIS and EIOS, time series suitable for anomaly detection by disease, and country/territory and region.
• To benchmark a range of anomaly detection algorithms across both WAHIS and EIOS time series.
• To evaluate the performance of these methods against the events observed, by disease, including their capacity to warn of the beginning of an epidemic wave, of the approach of its peak and of its post-peak decrease.
• To compare the anomalies derived from EIOS and from WAHIS, and to propose a combined alert rule exploiting the timeliness of open-source information and the reliability of official data.
• To consolidate and develop the analytical work into a reproducible, version-controlled and documented pipeline, designed to accommodate additional diseases, that WOAH staff can run, maintain, update and host on WOAH infrastructure.
• To design and develop dashboards and automated alerts presenting the early warning outputs, for internal WOAH use and, where relevant, for Members.

Missions and activities

Inception: material and methods

• A document describing the anomaly detection methods retained for the benchmark, with the rationale for their selection and their exclusion, building on the review of the literature made available by WOAH.
• A document describing the technical approach retained to consolidate and develop the pipeline, covering architecture, languages, version control, hosting, and the execution and update procedure.
• A document describing the development of dashboards and the alert products.
• A list of the data, accesses and infrastructure needed to apply the methodology selected.
• A work plan setting out the timeline for the deliverables.

Construction of time series from WAHIS and EIOS

• A characterisation of the WAHIS and EIOS data available for all the listed diseases, covering volume, granularity, and temporal and geographical coverage.
• A clean Masterfile of the WAHIS and EIOS time series, documented and reproducible.

Benchmarking of anomaly detection algorithms

• The implementation of the anomaly detection methods retained, applied to the WAHIS and EIOS time series for all the diseases.
• A comparative analysis of the performance of these methods by disease, and by stream of information.
• A recommendation on the anomaly detection strategy to be retained, stating whether a single generic method with disease-specific parameters or different methods according to the epidemiological profile should be used.

Consolidation and development of the pipeline

• A single version-controlled repository containing the consolidated code for the anomaly detection, the evaluation and the production of the alerts, with a documented structure allowing additional diseases to be integrated.
• A reproducible execution environment, with documented dependency management and run procedure, allowing the analysis to be re-run on updated data.
• The automation of the data update, of the execution of the models and of the publication of the results, with automated quality checks on the input data and on the outputs of the pipeline.
• Automated analytical reports, for example produced with Quarto, regenerated directly from the pipeline.
• A