Data Scientist

Education Qualifications

• Bachelor's or Master's degree in Computer Science, Data Science, Statistics (Statistical computing )or a related quantitative field.
  
• 3+ years of hands-on experience in machine learning and computer vision, with a demonstrable portfolio of deployed models.
  

Proficiency in:

• Python (PyTorch or TensorFlow) for deep learning model development.
  
• Object detection and image classification frameworks, particularly YOLO architectures (YOLOv8 or later).
  
• Data annotation tools and active learning workflows for building labeled datasets.
  
• Cloud platforms, specifically AWS, for model training, storage, and deployment.
  
• SQL and familiarity with data warehouse environments (Databricks preferred) for integrating model outputs with structured household data.
  
• Model deployment and MLOps practices, including CI/CD pipelines and experiment tracking with Weights & Biases or equivalent.
  
• Edge deployment optimization (TensorFlow Lite, ONNX) for low-connectivity field environments.
  

• Research, design, and implement image classification and object detection models (including YOLO-based architectures) for automated adoption t across RTV program domains including agriculture, WASH and livestock adoption practices.
  
• Build and maintain end-to-end ML training, validation, and test pipelines ensuring model accuracy, reliability, and generalizability to field conditions in low-resource environments.
  
• Optimize models for edge deployment in environments with limited connectivity, including TensorFlow Lite integration for mobile and offline use cases.
  
• Design and manage image data collection protocols and annotation workflows to produce high-quality labeled datasets for compliance indicator categories across all program domains.
  
• Integrate image metadata and classification outputs with the RTV data warehouse (Databricks medallion architecture) for correlation with household progression and adoption metrics.
  
• Develop automated adoption classification outputs that map to RTV's binary and weighted adoption scoring frameworks and validate against AHS survey-based assessments.
  
• Conduct structured experiments to benchmark model performance across deployment contexts (Uganda, Rwanda, DRC), applying Weights & Biases for experiment tracking and reproducibility.
  
• Build and document RESTful APIs to expose model predictions to WorkMate and other consuming field applications.
  
• Maintain clear documentation of model architectures, preprocessing pipelines, evaluation metrics, and versioning practices for cross-functional collaboration.