aida: a medical knowledge base for working doctors
A private library for working doctors, built so patient data has nowhere to go.
Open to Summer 2027 SWE internships in Singapore
hello@tuxedolphin.com
2025 · in progress · solo · react · typescript · supabase · cloudflare
I built aida’s schema so patient-identifiable data is awkward to store at all, behind 30 row-level-security policies.
A personal project, deployed. aida is a private library for doctors and medical students in Singapore: templates, snippets, and case reflections, searchable and browser-only. It is deliberately not a clinical decision tool and not a hospital record system. The interesting constraint was privacy by construction: there is no name field and no facility field anywhere in the schema, so the schema itself is the safeguard. The code is strict TypeScript with no escape hatches, the design decisions are written down as ADRs, and CI runs lint, types, unit and end-to-end tests, and a bundle-size budget on every push. It now runs at aida.tuxedolphin.com with hybrid search (384-dimension semantic similarity combined with Postgres full-text), row-level security on every row a doctor can see, and 1,112 Vitest tests plus Playwright end-to-end coverage in CI.
- data: no patient data
- types: strict, no any
- search: hybrid: pgvector 384-dim + full-text
- security: 30 RLS policies
- tests: 1,112 Vitest tests
try it
Back to Projects