European Commission, DG JRC - Joint Research Centre
Encore 11 jours pour répondre
data (description, +15), data engineering (description, +15)
machine learning (description, +15)
API (description, +15)
design (description, +15)
évaluation (description, +15)
Un signal dans l'intitulé vaut 40 points, une mention dans la description 15, jusqu'à 100. Le score d'un poste ne dit rien des autres : cet avis n'a pas de note générale.
A contract for research services to design, prototype and evaluate data pipelines for collecting, structuring, validating and monitoring evidence from heterogeneous public data sources. The pipelines should combine deterministic data engineering (API and database extraction, parsing), machine learning, and large language models (LLMs) with human expert validation. The contract will be structured in two broad packages. Work-package 1 supports INCITE (https://innovation-centre-for-industrial-transformation.ec.europa.eu). It shall develop a pilot to identify and update project-level information on emerging and established techniques for decarbonisation, depollution, resource efficiency and circular economy in industrial and livestock installations, covering technology, maturity, financing, performance and dependencies. Work-package 2 shall develop and tests methods to build and maintain a facility-level dataset of circular-economy installations in EU Member States. This covers facilities that recover, sort, process or recycle waste and by-products into secondary materials, including their location, capacities, technologies, ownership, inputs, outputs and commercial links. Both work-packages are exploratory pilots: the aim is to demonstrate and rigorously evaluate workflows, not to deliver a production platform. The contractor must deliver documented prototype pipelines, validated datasets, a quantitative evaluation of accuracy, coverage and cost, and a feasibility assessment for continuous updates and scale-up. In terms of the technical approach, the contractor is expected to propose a hybrid architecture, choosing the most appropriate technique for each task and justifying it on accuracy, cost, reproducibility and maintainability. For example: • Structured sources (e.g. EU funding databases such as CORDIS, Innovation Fund and LIFE; the EU Industrial Emissions Portal; national permit registries; Eurostat) can be accessed deterministically via APIs, bulk downloads or database extraction wherever possible. • Machine learning and rule-based methods can be used for classification, deduplication, geocoding and entity resolution across sources. • LLMs can be used mainly to extract structured information from unstructured documents (permits, company websites, press releases, scientific publications, sustainability and financial reports). Agent-based workflows can be used in cases where multi-step search is necessary.
Le dossier de consultation et les modalités de réponse sont sur l'avis officiel. Calloffre ne soumet jamais de réponse à votre place — l'outil vous aide à repérer les appels d'offres qui correspondent à votre profil et à préparer un premier brouillon de mémoire technique, que vous relisez et envoyez vous-même.
GUD - Umwelt- und Gesundheitsschutz (UGZ)
SPF Sécurité sociale - DG Personnes handicapées
SPF Sécurité Sociale
SPF Sécurité Sociale
Données issues du BOAMP (DILA), publiées sous Licence Ouverte / Etalab 2.0. L'avis officiel fait foi : Calloffre reproduit les informations de la notice mais ne garantit ni leur exhaustivité ni leur actualité.