Constellation

An explainable knowledge atlas for the world's rare diseases, built for the 7th Global AI Hackathon, Challenge 05: AI Atlas for the World's Rare Diseases (OpenAI × Buffalo Initiative × Hack-Nation).

For most rare diseases, the first clinician a child ever sees is not a specialist. It is a parent. Parents and caregivers hold the longest, richest clinical record that exists for their child, yet that record is scattered across discharge summaries, patient-portal messages, and memory.

Constellation is built for that first pediatrician, and for the scientists and advocates who can shorten the journey once the pieces are joined. Paste a note, and the atlas reads it: symptoms become ontology codes locally in the browser, the patient's profile lights up on a starfield map of mechanistic disease biology, and every connection shown (to a gene, a shared pathway, a patient community, a study, a researcher, or a repurposing candidate) carries its source, its confidence, and its contradictions. When the evidence runs out, Constellation says so plainly and shows which experiment would fill the gap.

The live prototype runs on a demo dataset and includes clinical note intake, an interactive mechanism map, a contradiction radar, a repurposing explorer, a one-click research dossier, and four persona journeys.