Kite Therapy : AI Matchmaking Engine for Pediatric Allied Health Services

Healthcare

Kite Therapy (Australia) – AI-Powered Therapy Matchmaking Platform

Kite Therapy partnered with us to solve one of Australia’s biggest challenges: the shortage of allied health professionals for children with developmental delays. We built a smart AI engine that matches children with the most suitable therapists based on diagnosis, location, therapy type, and therapist availability. The system automates intake forms, assesses compatibility, and schedules sessions, reducing the long waitlists that frustrate parents. Our AI also adapts over time as it learns which matches yield the best therapy outcomes. Kite Therapy now connects more families, more quickly—giving children faster access to the care they need.

Project Overview

Client: Kite Therapy (Australia-based pediatric therapy provider specializing in early intervention)
Challenge: Severe delays in connecting families with suitable allied health professionals
Goal: Build a smart AI engine to:
- Match children with compatible therapists based on needs, location, and availability
- Automate form intake, assessment, and scheduling
- Continuously improve matchmaking accuracy based on therapy outcomes
Team: 8 (3 ML Engineers, 2 Healthcare Data Analysts, 2 Backend Devs, 1 Clinical Product Manager)
Timeline: 5 months (AI Matching Prototype → Pilot with Clinics → Full Rollout)

“GenX helped us move from reactive scheduling to proactive care matching. The AI engine ensures every child is matched to the right therapist faster—without compromising clinical quality.”

Founder & Director, Kite Therapy

The Challenge

Critical Pain Points:

Families faced long waitlists, sometimes exceeding 6–9 months for therapy sessions
Manual therapist assignments lacked personalization and efficiency
No system existed to predict or improve long-term therapy match outcomes

Technical Hurdles:

Structuring intake data into actionable clinical profiles
Balancing hard constraints (e.g., diagnosis fit, travel radius) with soft preferences (e.g., therapist gender, communication style)
Using therapy success metrics and feedback to learn, without infringing on privacy.

Tech Stack

Component	Technologies
Matching Models	LightGBM, XGBoost, Python Rule Engines
Scheduling & Availability Engine	Firebase, PostgreSQL, Google Calendar API
Frontend Interfaces	React, Next.js, Tailwind
Intake Form Automation	NLP Parsers, OCR, React Hook Form
Data Privacy & Security	HIPAA-ready stack, AES-256 encryption, Australian Privacy Principles (APPs) compliance
Feedback & Outcome Analytics	Mixpanel, Looker Studio, Firebase Analytics

Key Innovations

AI matched children with therapists using diagnosis, location, and outcomes-based learning. Compatibility scores improved over time as the model learned from successful sessions. Waitlists shrank and access to care accelerated.

Diagnosis-to-Therapist Matching Logic

Mapped clinical input fields to allied health requirements and regional therapist filters

Result: 63% faster initial therapist assignment vs. previous manual workflows

Outcome-Aware Matching Engine

Learned over time which therapist-child pairings led to improved therapy retention and progress

Result: 41% decrease in no-shows and early dropouts

End-to-End Automation from Intake to Booking

Parents could complete all steps—discovery to first appointment—in under 10 minutes

Result: 72% improvement in conversion from inquiry to scheduled session

Our AI/ML Architecture

Core Models

Therapist Matching Engine:
- Gradient boosting + rule-based filters to match child profiles with therapist specialties and availability
- Incorporates clinical diagnosis, location, modality (speech, OT, physio), and therapy delivery mode (in-person/virtual)
Compatibility & Outcome Feedback Loop:
- Learns from past sessions to identify which variables influence long-term engagement and success
- Adjusts future match rankings using therapist ratings, cancellations, and outcome notes
Automated Intake & Scheduling System:
- Natural language parser for medical history fields
- Matches calendar openings and therapist types to session urgency

Data Pipeline

Sources
- Online intake forms, therapist schedules, diagnostic records, historical therapy outcomes
- Session feedback from families and clinicians

Processing: De-identified processing with encrypted storage and consent-based sharing mechanisms

Integration Layer

Custom web app for families and therapists
Syncs with clinic management tools and availability calendars (Google, HealthEngine, Cliniko)
Secure messaging and pre-visit intake forms with dynamic therapist recommendations