Data science internship – Clustering analysis of continuous glucose monitoring profiles in the conte

SFBI; Saint-Genès-Champanelle; On-site; EUR 8,900 - 11,000; Part time

Job summary

SFBI in Saint-Genès-Champanelle seeks a Master student for a 5–6 month data science internship focused on clustering analysis of continuous glucose monitoring (CGM) profiles within precision nutrition for healthy aging. The intern will preprocess CGM data, review CGM time series clustering literature, implement methods, and deliver a full clustering analysis with quality assessment and parameter tuning.

Work is in a multidisciplinary environment with experts in data science and nutrition

Qualifications

• Statistics and data analysis
• Programming in R or Python
• Dimension reduction and clustering on time series
• Experience with time series data or CGM is beneficial
• Literature review and scientific reporting
• Team-oriented multidisciplinary collaboration

Responsibilities

• Perform clustering analyses to identify glycemic subgroups from CGM profiles
• Preprocess CGM data for clustering analyses
• Survey literature on CGM time series clustering and select robust methods
• Implement chosen methods and provide a full clustering analysis with quality assessment

Skills

Statistics and data analysis

R

Python

Time series analysis

Clustering methods

Research methodology

Education

Master in Data Science or Statistics

Tools

R

Python

Job description

Data science internship – Clustering analysis of continuous glucose monitoring profiles in the conte

data science Clustering human health

Description

Data science internship – Clustering analysis of continuous glucose monitoring profiles in the context of precision nutrition for healthy aging

The intern will join the Human Nutrition Unit and more specifically the VaRIAe team (≈15 people), whose scientific objectives are to describe, understand, and predict inter-individual variability in metabolism and response to diet, with the ultimate goal of improving nutritional prevention of cardiometabolic diseases and supporting healthy aging.
Context and scientific background of the internship
Inter-individual variability in responses to food may be influenced by individual metabolic capacities, lifestyle and usual dietary habits. This internship is part of the MetabotypAGE clinical trial ( https://hal.inrae.fr/hal-05667287v2 ), an interventional nutritional study aiming to identify distinct metabolic profiles in non-diabetic older adults. During a 2-week free-living period, 130 participants carried a continuous glucose monitoring (CGM) device which continuously (15min interval) measures blood glucose concentration, a key biomarker of metabolic health. During this period, participants consumed four different standardized test meals at home. The aim of the internship is to test clustering methods applied to time series in order to identify subgroups of individuals with similar glycemic response profiles to test meals. Obtained results will allow the project’s next steps, consisting of subgroup characterization on multiple health dimensions, to better understand the biological underpinnings of group classifications.

Missions:

The intern will be responsible for clustering analyses that will produce the glycemic subgroups of interest. He/she will also preprocess CGM data before clustering analysis. The first step of the internship will be to make a study of the literature regarding CGM time series clustering to select the most robust approaches to be explored within the scope of this project, as well as quality criteria (robustness, validation …). Based on this first exploration, in collaboration with the project members and supervisors, an existing method or a subset of them will be chosen for implementation on the project data. The end goal will be to provide a full clustering analysis, including quality assessment of resulting clusters (e.g., robustness, silhouette) and optimization of the clustering method parameters.
This work will include methods based on classical clustering algorithms applied to time series extracted features, which can be generic or specifically crafted for CGM profiles. Methods based on distance metrics specialized for time series, like elastic distance combined with distance-based clustering methods, also falls in the scope. Depending on the profile of the intern, more complex methods can also be envisaged, including a mixture of mixed models, stochastic process modeling, or auto-encoders.
The intern fellow will be supervised by two experts respectively on data science and nutrition physiology and must therefore be comfortable with multidisciplinary scientific exchanges.

• Candidate Profile: last year of Master in Data Science or Statistics, with a strong interest in applying these disciplines to biology and health research within a multidisciplinary scientific environment. Candidates with a combined background in biology and statistics would also be highly valued.
• Statistics and data analysis
• Programming language for statistics and data analysis (R or python)
• Dimension reduction, unsupervised analysis and clustering methods experience on other projects
• Experience with time series data and/or CGM data would be beneficial
• Research methodology: bibliography, scientific reporting
• Candidates will work on a multi-disciplinary environment. It is expected from them to be open to a team work and interactions during regular project and team meetings.

Start date is adjustable, at the latest March 2027.
Duration: 5-6 months
Remuneration: in accordance with regulatory internship pay scales

Publications of the research group on the proposed topic

• Mosoni et al. Multidimensional phenotyping towards personalized preventive nutritional support for older adults, protocol for an open trial in >100 subjects. F1000. 2026. https://hal.inrae.fr/hal-05667287v2
• Lépine et al. Investigating the Postprandial Metabolome after Challenge Tests to Assess Metabolic Flexibility and Dysregulations Associated with Cardiometabolic Diseases. Nutrients. 2022 Jan 21;14(3):472. https://doi.org/10.3390/nu14030472