Le portail francophone de la géomatique

Contact : rsawahn@texte-a-enlever.umd.edu

Internship at NASA Harvest - University of Strasbourg
Keywords: Research, AI/ML, Software Engineering, Agronomy, Agriculture, Food Security

NASA Harvest is NASA’s global food security and agriculture consortium. Part of the core Harvest team is based at the University of Strasbourg, where we develop satellite-based agricultural monitoring and large-scale modeling systems that support global food security. The internship will be hosted jointly by NASA Harvest and the University of Strasbourg, and will be based at the ICUBE laboratory in Strasbourg.

Project Overview: Yield Prediction for Global Food Security
Accurate crop yield prediction is vital for sustainable agriculture, policy planning, and humanitarian decision-making. This is especially important in regions affected by conflict. As those often lack reliable ground observations, satellite-based modeling provides some of the only dependable information.
Our group develops and maintains a scalable yield prediction system (VeRCYe) that processes terabytes of geospatial data and uses biophysical crop simulation models alongside satellite observations. The internship contributes directly to improving and expanding this system.

Possible Internship Tasks
Based on the applicant’s interests and profile the tasks could be an adapted version of one of the following options:
1. Integration of New Crop Simulation Models: Integrate established crop models (such as DSSAT or WOFOST) to the existing pipeline. Develop and evaluate model ensembling approaches.

2. Crop Model Optimization: Explore optimization methods ranging from Reinforcement Learning to Bayesian Approaches for crop growth simulators (models) for identifying optimal cultivars, or optimization of cultivars parameters for different regions of interest (all within simulation environments).

3. Soil Data Integration: Build a pipeline to fetch soil variables from global soil databases for an automatic Integration of soil information into current workflows.

4. Benchmarking Meteorological Data Sources: Compare and assess multiple weather data sources (such as ERA5, NASA-POWER, and CHIRPS) and quantify their impact on yield prediction accuracy.

5. Full-Stack System Development: Help generalize and modularize the existing yield prediction system (FastAPI/React). As part of our team is also affiliated with Microsoft AI for Good Lab, transitioning existing models from HPC to Azure Cloud can also be explored.

Candidate Profile
We welcome applicants from computer science, agronomy, AI/ML, remote sensing, environmental science, or related fields. Proficiency in Python and fluency in English (written and spoken) are required. Everything else can be learned. Motivation and willingness to explore are most important.
Location, Hours, and Duration
·    Location: ICUBE, University of Strasbourg (hybrid options possible)
·    Hours: 35 hours per week
·    Duration: Typically 3–6 months, with flexibility
·    Gratification: 662€/month. Additionally, applicants studying in Erasmus+ states may be eligible for Erasmus+ funding (~+650€/m depending on their home institution’s policies).
·    Depending on the topics chosen, we are happy to aim for a publication if of interest.

Applications and Contact:
Applications will be reviewed on a rolling basis as they are received. The position will remain open until filled or until 31 January 2026. Early application is strongly encouraged, as a security clearance is required and this may take some time. Please submit your CV and cover letter (both in English) to rsawahn@texte-a-enlever.umd.edu . If you have any questions, don’t hesitate to just reach out under the same address. We will be happy to discuss any ideas and concerns.