A clinical documentation platform
An orthopaedic surgery practice in Australia wrote its operative notes by hand. I built the platform that turns one structured form into the note, the referral letter and clean research data.
The problem
Surgeons wrote operative notes by hand, and research data sat in three systems that did not talk to each other.
The first prototype kept every form answer in one large JSON field. That blocked research queries, partial imports and an audit trail.
What I built
Built a multi-tenant platform in React, TypeScript, Fastify and PostgreSQL on AWS, with sign-in and role checks.
Designed a structured encounter form with 2,297 fields, with templates per surgeon and a clinic default behind them.
Built a document pipeline from Word templates that writes the operative note and the referral letter.
Moved the data from one JSON field to one row per answer with no downtime, so research queries and partial imports work.
Isolated each practice with row-level security in PostgreSQL.
Connected the platform to REDCap, Dataverse and a document generation service.
How the work ran
The work ran over 11 months. Each handover came with a written report and runbooks for the parts it touched.
Result
The form first rendered about 15,000 page elements in 8 to 12 seconds. It now renders about 500 in under a second.
A save that reported success but wrote nothing came from six separate causes. I traced and fixed all six.
The final handover had a report of 17 sections, runbooks for each subsystem, a test suite and a local setup that runs in under 15 minutes.
The practice, its surgeons and the product name are left out. The load figures come from my own development report, not an outside audit.
