Machine Learning consultant

Machine Learning Consultant – Data Science for Programmes: Needs Assessment Clustering and Predictive Modelling Pilot

The International Rescue Committee (IRC) is seeking a solutions-driven, passionate change-maker to join their team as a Machine Learning Consultant. This role is crucial for leveraging data science to improve humanitarian aid delivery in crisis-affected regions.

About the International Rescue Committee

The IRC responds to the world's worst humanitarian crises, aiding in the restoration of health, safety, education, economic wellbeing, and power to individuals impacted by conflict and disaster. Operating globally, the IRC works at the intersection of humanitarian response and long-term development. The METAL Unit within the IRC focuses on monitoring, evaluation, technology, accountability, and learning, utilizing advanced analytics to derive actionable insights from programme data.

Background

Building on a successful machine learning pilot in Ecuador-Peru, where clustering identified underserved populations, the IRC aims to expand this initiative. This consultancy will focus on applying clustering for needs assessment in a new context and developing a predictive model for a specific programme intervention. This work is part of the broader ML for Programmes initiative.

Objectives

The Consultant Will:

• Apply unsupervised machine learning (clustering) to client and population data from [Country A] to identify distinct population segments for needs assessment.
• Conduct a methodological assessment and attempt to apply Multilevel Regression and Post-stratification (MRP) techniques for estimating needs in data-sparse areas, reporting findings and confidence levels.
• Conduct a data feasibility assessment for predictive modelling in [Country B], identifying the most suitable pilot programme.
• Produce a methodological guide for IRC Regional Measurement Advisers (RMAs) to assess data readiness, commission analyses, and communicate findings.
• Build, validate, and document a predictive model for the selected programme, adhering to best practices for development, robustness testing, and responsible deployment.

Scope of Work

Workstream A: Needs Assessment Clustering – [Country A]

A1 Data Assessment and Preparation.

Review available datasets (client registration, MSNA, programme monitoring), assess quality and suitability for clustering, and document findings in a brief data assessment note. Conduct necessary data cleaning and structuring.

A2 Clustering Analysis.

Apply appropriate clustering algorithms (e.g., K-medoids, hierarchical clustering) to identify population segments. Assess and attempt MRP techniques for data-sparse locations, reporting estimates with uncertainty intervals and limitations.

A3 Needs Assessment Report.

Produce a report interpreting cluster outputs and MRP estimates in programmatic terms, translating statistical findings into actionable insights for programme design and targeting.

Workstream B: Predictive Modelling Pilot – [Country B]

B1 Data Feasibility Assessment.

Review candidate programme datasets, assess suitability for predictive modelling based on outcome data, sample size, completeness, and ethical considerations. Recommend a pilot programme.

B2 Model Development and Validation.

Build a predictive model using historical programme data, following best practices including documented train/test splits or cross-validation, robustness checks, feature importance analysis, and clear documentation of limitations and retraining conditions. Analytical code to be provided in R or Python.

B3 Operational Tool and Documentation.

Develop a lightweight operational tool (e.g., scoring widget, dashboard integration) for programme staff, with accompanying documentation on its use, interpretation, limitations, and human oversight protocols. Deployment within existing Databricks and Power BI infrastructure.

Workstream C: Methodological Guide for Regional Measurement Advisers

C1 RMA Methodology Guide

Produce a practical methodological guide (8–10 pages) for RMAs on assessing data suitability, selecting methods, commissioning analyses, checking outputs, and presenting findings, grounded in consultancy experience.

Deliverables and Timeline

• Deliverable 1: Data assessment note (Workstream A) - Weeks 1–2 (Est 3 days)
• Deliverable 2: Data cleaning and preparation (Workstream A) - Weeks 2–5 (Est 7 days)
• Deliverable 3: Clustering analysis and MRP assessment (Workstream A) - Weeks 5–9 (Est 8 days)
• Deliverable 4: Needs assessment report, final (Workstream A) - Weeks 9–12 (Est 5 days)
• Deliverable 5: Data feasibility assessment note (Workstream B) - Weeks 2–4 (Est 3 days)
• Deliverable 6: Model development and validation (Workstream B) - Weeks 8–14 (Est 8 days)
• Deliverable 7: Operational tool and documentation (Workstream B) - Weeks 14–17 (Est 4 days)
• Deliverable 8: RMA methodology guide (Workstream C) - Weeks 15–18 (Est 4 days)

Total Estimated Effort: 42 days. Workstreams A and B run in parallel. Assumes prompt data access and sign-offs.

Ethical Considerations and Data Protection

The consultant must sign a data sharing agreement and adhere to the IRC's data protection policy. All provided data will be de-identified. Predictive models are for resource allocation and tailored support, not automated decision-making affecting individual cases. Robustness checks and subgroup performance analysis are mandatory. Ethical parameters must be reflected in all outputs.

Required Qualifications

• Advanced degree (Master's or PhD) in a quantitative field (economics, data science, statistics, computer science) or equivalent professional experience.
• Demonstrated experience in clustering and predictive modelling with real-world datasets.
• Proficiency in R or Python; experience with Databricks is a plus.
• Proven experience with messy, incomplete, or administratively-generated data.
• Experience with MRP or similar small-area estimation techniques is desirable.
• Familiarity with the humanitarian or international development sector is ideal.
• Strong technical writing skills for non-specialist audiences.

Management and Reporting

The consultant will report to [Philip Blue / Regional Measurement Adviser, Latin America] and collaborate with the IRC's Measurement Unit and country programme MEAL staff. An inception call, fortnightly check-ins, and two formal review points are scheduled. The methodology guide will undergo review by an RMA.

Submission Requirements

• CV highlighting relevant data science experience.
• Short expression of interest (max one page) detailing approach to one workstream.
• Two examples of prior analytical work (outputs, code repositories, or reports).
• Daily rate in USD.
• Contact details for two referees.

Professional Standards

All IRC workers must adhere to the IRC Way - Standards for Professional Conduct: Integrity, Service, Equality, and Accountability. This includes adherence to policies on Safeguarding, Conflicts of Interest, Fiscal Integrity, and Reporting Wrongdoing. The IRC is committed to creating a safe environment and taking corrective actions when harm occurs. IRC values critical reflection, power sharing, debate, and objectivity.

Key skills/competency

• Machine Learning
• Data Science
• Clustering
• Predictive Modelling
• R
• Python
• MRP
• Data Analysis
• Quantitative Research
• Humanitarian Sector