Sophia Foundation Photo-Matching Algorithm
RootSquare built a multimodal AI system for the Sophia Foundation for Children to automatically match children’s photos with fragmented records. The solution reduced days of manual work to minutes while keeping humans in the loop for verification and trust.
Sophia Foundation for Children
RootSquare partnered with the Sophia Foundation for Children to automate the organization, verification, and matching of child records across images and fragmented documentation.
The Foundation operates in environments where records are often incomplete, inconsistent, and manually maintained. Photos, names, and background information frequently arrive separately, making identification slow and error-prone. The process required extensive manual review and carried high risk of duplication or mislabeling.
We designed a modular, multimodal AI system that processes both images and scanned documents, extracts structured information, and determines whether a child already exists in the database or represents a new record. Matching prioritizes face-based similarity and falls back to full-image analysis when needed.
Ambiguity is not hidden. When the system detects uncertainty, it flags the case for human review. This ensures accuracy while dramatically reducing manual workload.
Tasks that previously required days of visual inspection can now be completed in minutes, while maintaining transparency and auditability.
Images used in public demonstrations are synthetic and contain fictional information to protect privacy. Dashboard visuals are authentic and shared with the foundation’s permission.
This project reflects our core philosophy: AI should not replace human judgment. It should strengthen it through structured, explainable automation built for real-world constraints.
