ToR: Remote Sensing and Machine Learning Consultant – Infrastructure Monitoring Module, eEARTH

1. Background

Crisis Group is developing an infrastructure monitoring module within the eEARTH interface, a broader platform focused on the analysis of climate, resource and conflict-related dynamics. The module aims to support the monitoring of critical infrastructure developments and changes over time using satellite imagery. The project will primarily rely on publicly available earth observation data, particularly Sentinel-1 (SAR) and Sentinel-2 (optical) imagery, to detect and analyze changes such as infrastructure damage, (re-)construction, and expansion (e.g., ports, roads, buildings).

The module is intended to support frequent monitoring across selected geographies and should be designed with scalability, usability and analytical relevance in mind.

2. Objective of the Consultancy

The objective of this consultancy is to provide technical backstopping and expert guidance to Crisis Group’s team during the development of the AI-based system for detecting infrastructure changes from satellite data.

The consultant will advise on appropriate data selection and machine learning approaches for infrastructure change detection, review methodological choices, troubleshoot technical issues, and provide feedback on interim and final outputs.

3. Scope of Work

The consultant will provide targeted, short-term technical support (estimated at 15-24 hours in total) across the following areas:

3.1 Methodological Guidance and Model Selection

Advise on suitable AI/ML approaches for change detection using satellite imagery, including:
- Classical machine learning vs. deep learning approaches
- Supervised vs. unsupervised methods
- Time-series analysis techniques
- Practical trade-offs between model accuracy, interpretability, computational cost, and scalability

Provide recommendations on handling multi-modal data (e.g., combining Sentinel-1 and Sentinel-2)
Suggest best practices for preprocessing, feature extraction, and data labeling

3.2 System Design for Continuous Monitoring

Provide guidance on designing a scalable and efficient monitoring pipeline, including:
- Data ingestion and updating workflows
- Automation and scheduling for frequent monitoring
- Strategies for generalizing across different geographies

● Recommend best practices for maintaining model performance over time (e.g., retraining, validation)

3.3 Ad Hoc Technical Support

Provide troubleshooting support during development
Advise on addressing technical challenges as they arise
Review interim outputs and provide feedback

3.4 Final Evaluation

Review the implemented system once operational
Provide an expert assessment of:
- Model performance and robustness
- Scalability across geographies
- Sustainability of the workflow

4. Deliverables

The consultant will provide the following deliverables:

Initial consultation and written summary of technical recommendations on model selection and system design;
Ongoing advisory support during development (as needed within the agreed level of effort);
Feedback on selected interim outputs produced by the project team;
Final technical review note summarizing key findings, limitations and recommendations.

5. Duration and Level of Effort

The total estimated level of effort is 15-24 hours.

The consultancy is expected to start around July 2026 and be completed by late 2026 or early 2027, depending on project progress.

The work will be conducted remotely, with inputs provided through virtual meetings, email exchanges and written comments on selected outputs.

Requirements

6. Required Qualifications

The consultant should have:

Demonstrated expertise in machine learning and/or deep learning applied to geospatial or remote sensing data;
Experience working with satellite imagery, including Sentinel-1, Sentinel-2, or comparable datasets;
Familiarity with change detection techniques and time-series analysis;
Experience designing scalable data pipelines and monitoring systems
Ability to assess methodological trade-offs, strong problem-solving and experience providing technical advisory support to multidisciplinary teams.

1. Background

2. Objective of the Consultancy

3. Scope of Work

Supervised vs. unsupervised methods

Time-series analysis techniques

Data ingestion and updating workflows

Provide an expert assessment of:

Model performance and robustness

Scalability across geographies

Sustainability of the workflow

4. Deliverables

5. Duration and Level of Effort

Requirements

6. Required Qualifications

The consultant should have:

7. Reporting and

Consultant, Early Action & Risk Tracking Hub (eEARTH) at International Crisis Group