2-year Post-Doctoral Research Visit: Inria, France

Summary

Inria is inviting applications for a 2-year Post-Doctoral Research Visit within the eTaxonomist ANR JCJC project (2026–2030). The project aims to advance biodiversity monitoring by moving beyond standard “black-box” deep learning models to develop computer vision and NLP systems that emulate expert, trait-based reasoning (e.g., leaf shape, beak curvature, wing patterns). The post-doctoral researcher will focus on building automated pipelines that convert unstructured expert domain knowledge (floras, field guides, web data) into structured, machine-readable knowledge bases and linking them to visually grounded trait models in collaboration with platforms like Pl@ntNet.

Post-Doctoral Research Visit: Trait-Based Species Identification via Knowledge Extraction and Weakly Supervised Learning

Key Information Overview

Research Area

• Theme / Domain: Data and Knowledge Representation and Processing; Biology, Health, and Earth/Life Sciences (BAP A)
• Fields: Natural Language Processing (NLP), Computer Vision, Knowledge Representation / Ontologies, Machine Learning, Biodiversity Informatics

Location

• City: Montpellier, France
• Affiliation: EVERGREEN Team, Centre Inria d’Université Côte d’Azur (supervised in collaboration with Diego Marcos, Alexis Joly, and Zeynep Akata from TU Munich).

Eligibility & Qualifications

Required Qualifications & Skills:

• Degree: PhD in Natural Language Processing, Knowledge Representation, Computer Vision, or a closely related area of Machine Learning / Computer Science / Applied Mathematics.
• Programming & Frameworks: Strong programming skills in Python and hands-on experience with deep learning frameworks (PyTorch).
• LLM Expertise: Demonstrated experience working with Large Language Models (prompting, in-context learning, self-supervised fine-tuning).
• Collaboration & Language: Ability to work independently and collaboratively in an interdisciplinary environment; proficiency in written and spoken English.

Desirable / Appreciated Skills:

• Experience with knowledge graphs, ontologies, or structured knowledge extraction.
• Familiarity with vision-language models (e.g., CLIP) or weakly supervised visual representation learning.
• Interest or prior experience in ecology, biodiversity, or natural history applications.
• Experience with large-scale High-Performance Computing (HPC) environments (e.g., Jean Zay).

Job Description & Main Activities

The selected candidate will contribute primarily to WP1 (Domain Knowledge Base Creation) and WP2 (Visually Grounded Trait-Based Descriptions):

• Corpus Curation: Assemble and curate heterogeneous textual corpora of morphological species descriptions covering plants, insects, and birds.
• Pipeline Design: Design and evaluate LLM-based pipelines for domain ontology construction, utilizing in-context learning and domain-adapted fine-tuning strategies.
• Knowledge Base Population: Populate the knowledge base with structured tuples (class, entity, quality, value) and comparative facts by integrating LLM outputs with existing biodiversity databases (TRY, GBIF, eBird, EOL TraitBank).
• Computer Vision Integration: Collaborate with project PhD students on weakly supervised computer vision methods for part-aware representation learning and trait prediction from images.
• Evaluation & Dissemination: Set up rigorous evaluation protocols (precision/recall against expert-curated gold standards, downstream zero-shot utility) and contribute to peer-reviewed publications and open-source software/data releases.

Benefits Package

• Leave: 7 weeks of annual leave + 10 extra RTT days off + exceptional leave options.
• Work Flexibility: Subsidized meals, partial public transport reimbursement, teleworking options (after 6 months), and flexible working hours.
• Equipment & Social: Professional IT/videoconferencing equipment provided, social/sports activities, access to vocational training, and comprehensive French social security coverage.

How to Apply

1. Online Submission: Applications must be submitted directly through the official Inria Job Portal (Offer #2026-10382).
2. Registration: Enter your email address on the portal to create/save your candidate profile and submit supporting documents (CV, cover letter, references, etc.).
3. Contact for Queries: Diego Marcos Gonzalez (diego.marcos@inria.fr).

Last Date for Application

• Deadline: September 18, 2026

Link