Postdoctoral position in frugal AI for bioacoustic analysis of natural environments (M/F)

Job Description

Missions

The person recruited will have as their main mission to pursue and consolidate a research project on frugal AI for the bioacoustic analysis of natural environments, within the IMAGINE team at LIRIS. The position aims to turn a promising exploratory effort into a structured scientific programme combining machine learning, representation learning, the analysis of sensor-based audio data, and biodiversity monitoring.
The person recruited will play a central scientific role in the design, training, evaluation, and adaptation of deep learning models applied to bioacoustic data. In particular, the work will involve developing generic representations capable of overcoming the limitations of current models, which are often focused on birds, poorly suited to rare or underrepresented species, and require large volumes of annotated data.
The mission also has a strong applied dimension: applying the models developed to concrete problems in acoustic ecology, particularly using real soundscape data, in order to contribute to the automatic analysis of biodiversity. The work should lead to a demonstrator illustrating the full processing chain, from representation learning to exploitation on real bioacoustic data.
Finally, the postdoctoral researcher will play a driving role in the scientific structuring of a future collaborative project, with partners in ecology, bioacoustics, computer science, and environmental sciences.

Activity

Main activities:
- Consolidate the bioacoustic representation model. The person recruited will further develop the existing methodological work in order to produce a model capable of learning relevant representations from varied bioacoustic data, not limited to birds.
- Model design and experimentation: train and evaluate models. They will implement training protocols, select relevant datasets, define evaluation metrics, compare model performance, and analyse the results.
- Develop learning approaches suited to limited volumes of annotated data. The position will involve working on representation learning, self-supervised learning, few-shot learning, fine-tuning, and knowledge distillation, in order to make the models adaptable to new species, taxa, or ecological contexts.
- Apply the models to real bioacoustic data. The person recruited will use soundscape data to test the generality, robustness, and usefulness of the models on concrete ecological problems.
- Produce a software demonstrator. The work should result in a demonstrator illustrating the full pipeline: data preparation or adaptation, representation learning, adaptation to a target task, automatic analysis, and presentation of the results.
- Contribute to scientific output. Depending on the results obtained, the person recruited will contribute to the writing of scientific articles focusing either on methodological contributions in machine learning or on ecological applications.
- Contribute to the structuring of a national consortium. The person recruited will play an active role in the scientific preparation of a collaborative project with national partners.

Secondary activities:
- Take part in interdisciplinary exchanges. The person recruited will contribute to meetings, discussions, and workshops with partners in ecology, bioacoustics, artificial intelligence, and environmental sciences.
- Follow the state of the art. The person recruited will carry out scientific monitoring of audio foundation models, self-supervised learning methods, audio-language approaches, bioacoustic datasets, and associated evaluation methods.
- Contribute to the exploration of other funding schemes. Depending on the progress of the project, the person recruited may take part in discussions on European or international collaborations.

Your Profil

Skills

- PhD in computer science, with a focus on signal processing, image processing, or machine learning. Strong skills in machine learning and deep learning.
Proficiency in representation learning, self-supervised learning, fine-tuning, knowledge distillation and, ideally, few-shot learning methods.
- Experience in processing complex sensor-based data, ideally audio or time-series data.
- Excellent command of scientific programming, particularly in standard deep learning environments.
- Ability to read, understand and synthesise the international state of the art on audio models, computational bioacoustics and frugal AI.
- Ability to interact with non-computer-science partners, in particular ecologists or biodiversity specialists.
- Autonomy in conducting a research project, from methodological design to scientific dissemination.
- Strong writing skills for the production of scientific articles and reports.
- Ability to contribute to the structuring of a scientific consortium and to the preparation of a national collaborative project.

Your Work Environment

The position will be based within the IMAGINE team at LIRIS, as part of a project at the intersection of pattern recognition, machine learning, sensor-based data analysis and bioacoustics. It extends the team's scientific priorities towards a new modality, bioacoustic audio, and towards an application area with strong environmental impact: biodiversity monitoring.

Constraints and risks

The person recruited will be based at the LIRIS premises on the LyonTech-la Doua campus in Villeurbanne.

Compensation and benefits

Compensation

From €3,072 gross per month, depending on experience and the applicable CNRS salary scale.

Annual leave and RTT

44 jours

Remote Working practice and compensation

Pratique et indemnisation du TT

Transport

Prise en charge à 75% du coût et forfait mobilité durable jusqu’à 300€

About the offer

About the CNRS

The CNRS is a major player in fundamental research on a global scale. The CNRS is the only French organization active in all scientific fields. Its unique position as a multi-specialist allows it to bring together different disciplines to address the most important challenges of the contemporary world, in connection with the actors of change.

CNRS

The research professions