Machine Learning Engineering Consultant, Data and Analytics, DAPM, NYHQ, remote. Req

UNICEF is the world’s leading children’s rights organization, dedicated to promoting children’s survival, protection, and development. The Machine Learning Engineering Consultant will work in the Data and Analytics Section to support projects related to machine learning and artificial intelligence, enhancing data analysis and information extraction methodologies.

• Maintain and optimize the vaccine stockout machine learning model already trained for UNICEF’s Program Group Immunization Division, ensuring its accuracy, performance, and sustainability.
• Continue to enhance methods for large-scale data and information extraction from diverse and unstructured document sources for the West Central Africa Region Social Policy teams and the WASH Analytics team before moving to additional domains.
• Support the development of automated briefs and reports generation pipelines.
• Test, evaluate, and implement robust frameworks for (semi)-automated GenAI report and data quality assurance.
• Contribute to geospatial (GIS) and AI‑related initiatives, particularly as part of the Frontier Data Network Ahead of the Storm project.
• Provide technical advice on AI/ML approaches.
• Build reproducible workflows and contribute to machine learning and GenAI knowledge transfer within the team.
• Maintain, retrain, and document existing ML models in production when new data or new features are available.
• AI assisted data and information extraction from unstructured documents.
• Prototypes and production‑ready solutions for automated reporting.
• Contributions to GIS and AI project outputs.
• Define automated, AI‑driven data tests, also over GenAI outputs.
• Support Data and Analytics Section development of GenAI Retrieval Augmented Generation (RAG) driven SDG Country Briefs.
• RAG system that generates SDG country briefs contextualized per country with actionable information that helps countries understand where they are and are not meeting their SDG targets.
• Support UNICEF teams in extracting budget lines from governments’ public budget documents (particularly child‑related budget items) and data from unstructured documents for West Central Africa, ensuring the tool is adaptable for extracting information in any domain.
• A generic data extraction pipeline using the most appropriate technologies (e.g., Data Bricks, Microsoft Document Intelligence and custom machine learning models) that retrieves and structures relevant data and information from unstructured budget documents so that non‑technical users can query and retrieve the structured data they require to analyse and track national spending on child‑related budget items, including time‑series.
• Develop and deploy a user‑friendly interface for non‑technical programme teams to work with the document extraction tool for extracting information on child‑related budget items.
• Data Bricks/Microsoft Document Intelligence data extraction pipeline equipped with a user‑friendly interface to extract information on child‑related expenditures from unstructured national budget documents, including dynamic links to sources of every number, LLM‑generated analysis, time‑series, and charts.
• Coordinate with stakeholders (PG‑I, DAPM) to support maintenance of the released vaccine stockout model, retraining as new data sources become available and assisting in deployment as required.
• Vaccine stockout model maintained and integrated with the existing data pipeline producing predictions consumable by Power BI and other systems for the current set of 30 countries.
• Automated AI‑driven data and RAG‑generated report QA – Research phase?
• Test and report results from multiple evaluation frameworks, including but not limited to Deep Eval, Lang Chain Evaluation, Tru Lens, and RAGAS.
• Automated AI‑driven data and RAG‑generated report QA – Implementation phase?
• Implement into the production lifecycle and document the evaluation framework(s) that deliver the best results for the SDG Country Briefs and Document Extraction products.
• Work with D&A Geospatial lead to integrate the use of satellite embedding datasets to drive new geospatial analytics in our enterprise geospatial and data platforms.
• Geospatial foundation models…