The Challenge

Occupational Therapists working with Australia’s National Disability Insurance Scheme (NDIS) participants were spending between 16 and 20 hours writing, compiling, and formatting a single participant report. Because NDIS primarily funds consultation time and travel, these extensive documentation hours represented unpaid, unbillable operational burden.

The Choice

While leveraging a hosted cloud AI service would have been faster and cheaper to set up, it meant sending sensitive participant clinical data off-site. We advised against cloud models to protect patient privacy and recommended deploying an open-weight large language model (LLM) locally on their own GPU infrastructure. Even though this was our first time deploying LLMs on dedicated GPU environments, we chose data privacy and local ownership over the easier cloud alternative.

What We Built

We engineered an end-to-end, privacy-first AI report synthesis platform:

  • Self-Hosted LLM Engine: Deployed an open-weight large language model on local GPU hardware running Python backend services.
  • Therapist Web Portal: Built a React frontend where therapists input raw notes and generate structured draft reports in under an hour.
  • Client-Side Field-Level Encryption: Secured participant data within a document database with client-side field-level encryption.
  • Custom Prompt Pipeline: Systematically tuned prompts based on historical clinical records to consistently mirror official NDIS reporting standards.

The Outcomes

  • Tech Stack: Python, React, Open-Weight Large Language Model (LLM), Document Database, Dedicated GPU Hardware.
  • Compliance & Data Protection: Built to strict Australian healthcare safeguards and HIPAA privacy standards; participant data remains fully encrypted at rest using client-side field-level encryption.
  • Impact & Performance: Reduced report generation time by 95% (from 16–20 hours to ~1 hour).
  • Support & Partnership: Delivered 3+ months of extended post-go-live support at no extra cost to tune the model and ensure adoption—reflecting Rite Technologies' standard approach, where most of our partners rely on us for between 5 and 10 years across their evolving software lifecycles.

“Your leadership was a cornerstone of this success. The clarity, direction, and purpose you bring to your team are truly inspiring.”

— Head of Operations & Technologies

The Handover

By keeping participant data on-premises and fine-tuning an AI engine locally, we eliminated hundreds of hours of manual administrative burden for therapists while maintaining absolute data privacy. We continue to support their team whenever operational or infrastructure updates arise.

Tech Stack

LLM
On Premises AI
Python
React
GPU
Document Database
Field Level Encryption
HIPAA
NDIS
Health Tech
Occupational Therapy

If your organization needs AI-driven automation while keeping sensitive clinical notes entirely on-premises,

let's talk