Postdoc: Trait-Based Species Identification & Knowledge Extraction

Inria; Montpellier; Sur place; EUR 42 000 - 54 000; Plein temps

Avantages offerts par ce poste

Transport refund

Annual leave

Telework after 6 months

Equipment provided

Social events

Résumé du poste

Inria seeks a post-doctoral researcher to advance trait-based species identification through knowledge extraction and weakly supervised learning. The project targets automated, interpretable trait reasoning across plants, insects, and birds, using structured knowledge bases and visually grounded descriptions.

The role emphasizes building pipelines to convert expert texts into machine-readable knowledge and grounding traits in images, within a collaborative, interdisciplinary team at Inria Côte

Qualifications

• PhD in NLP, Knowledge Representation, computer vision, or ML.
• Strong Python programming and deep learning experience (PyTorch).
• Experience with large language models (prompting, in-context learning, fine-tuning).
• Excellent written and spoken English.
• Proven publication record in top CS venues.

Responsabilités

• Assemble and curate heterogeneous textual corpora across plants, insects, and birds.
• Design and evaluate LLM-based pipelines for domain ontology construction.
• Populate knowledge bases with structured tuples and integrate external databases.
• Contribute to weakly supervised vision methods for trait prediction from images.
• Establish evaluation protocols with domain experts and publish results.

Connaissances

Python

PyTorch

LLMs

Prompting

Independent worker

Publications

English

Formation

PhD in NLP / Knowledge Representation / CV / ML

Description du poste

Inria seeks a post-doctoral researcher to advance trait-based species identification through knowledge extraction and weakly supervised learning. The project targets automated, interpretable trait reasoning across plants, insects, and birds, using structured knowledge bases and visually grounded descriptions.

The role emphasizes building pipelines to convert expert texts into machine-readable knowledge and grounding traits in images, within a collaborative, interdisciplinary team at Inria Côte