KuraPath
know your health
Built for the Future of Health Education AI.
Defensible deep tech IP. Production-grade infrastructure. No prototypes.
KuraPath is built on eight deep tech capabilities that compound into a health education infrastructure moat. Every system described here is live in production and serving real users. This is not a roadmap or a research paper. It is shipped deep tech software powering the next generation of health education.
On-Device Health Education AI
Privacy-first AI that never leaves the browser
Predictive Health Education Trajectories
Learn where your biomarkers are trending
Privacy-Preserving Population Insights
Learn from everyone, expose no one
Intelligent Document Understanding
Any health document, understood in seconds
Synthetic Health Data Engine
Realistic test data at scale, zero privacy risk
Digital Health Twin
Your health, explored
Proprietary LLM Infrastructure
Provider abstraction + local AI inference
Biosphere World Model
Causal Bayesian network for environmental health risks
Satellite Environmental Intelligence
10m-resolution Earth observation for health risk
Eight capabilities. One compounding moat.
Each capability is a standalone technical achievement. Together, they create a compound advantage that is extraordinarily difficult to replicate. This is deep tech infrastructure, not just features.
On-Device Health Education AI
Privacy-first AI that never leaves the browser
Most health AI platforms send sensitive data to cloud servers for processing. KuraPath runs inference directly in the browser, with zero data transmission to external endpoints. Our dual-backend architecture supports both cloud and on-device modes, with users able to opt into a fully private mode where nothing ever leaves their device. Building production-grade, on-device health education AI is one of the hardest problems in health tech. We have shipped it.
On-Device Inference
WebGPU · Privacy Mode
Data Transmitted
0 bytes
100% On-Device
All inference runs locally via WebGPU. No data leaves the browser.
🔒 Zero Cloud Dependency
Predictive Health Education Trajectories
Learn where your biomarkers are trending
Traditional health education is static: you get a blood test, see a number, and look up what it means. KuraPath provides educational trend analysis showing where your biomarkers are trending based on your history. Our ensemble approach combines statistical models with contextual analysis to generate educational trajectory visualisations with confidence intervals. We cross-correlate biomarker trends with environmental exposure data (air quality, UV, pollutants), creating an educational context layer that no other consumer platform offers. This is not a diagnostic tool; it helps users understand their health trends to have more informed conversations with their doctor.
Biomarker Forecasting
Ensemble Prediction Engine
Biomarkers Tracked
37
HbA1c projected to normalise within 8 weeks at current trajectory.
📈 94% Confidence
Privacy-Preserving Population Insights
Learn from everyone, expose no one
Health education improves when informed by population-level patterns, but privacy regulations prevent data sharing across institutions. KuraPath addresses this with federated learning and differential privacy: the model learns from population patterns without any raw data ever leaving a user's device. Only anonymous, mathematically noisy weight updates are aggregated. The result is richer, population-informed health education content with formally guaranteed individual privacy.
Federated Learning
Differential Privacy
Privacy Budget
ε = 5.0
Zero raw data leaves any device. Only noisy weight updates are aggregated.
🔐 Mathematically Private
Intelligent Document Understanding
Any health document, understood in seconds
Health documents come in dozens of formats, from handwritten GP notes to complex pathology PDFs. KuraPath deploys 7 specialised extraction pipelines with clinical named entity recognition trained on Australian coding systems: ICD-10-AM, PBS, SNOMED CT, and ACHI. We extract, normalise, and structure document content that competitors cannot even parse, then translate it into clear, plain-language health education for the user. Our AU-specific clinical coding depth is a significant competitive differentiator in the Australian and APAC markets.
Document Intelligence
Clinical NER Pipeline
Active Pipelines
7
Clinical NER trained on AU coding systems. 7 specialised extraction pipelines.
📄 ICD-10-AM · PBS · SNOMED
Synthetic Health Data Engine
Realistic test data at scale, zero privacy risk
Sharing real health data for development, testing, or research is nearly impossible due to privacy constraints. KuraPath generates clinically accurate synthetic health profiles across 44 biomarkers and 74 languages using parametric generation. Every synthetic dataset is stress-tested against 9 adversarial attack vectors to ensure it cannot be reverse-engineered to reveal real patient data. Built specifically for CRO and pharmaceutical safety testing workflows where realistic data is essential but real data is off-limits.
Synthetic Data Engine
Parametric Generation
Coverage
37 Bio × 72 Lang
9 adversarial vectors tested per batch. Zero reverse-engineering risk.
🧪 All Tests Passed
Digital Health Twin
Your health, explored
Most people have no way to understand how lifestyle choices relate to their health profile. KuraPath builds an educational digital twin using a Bayesian causal network with Monte Carlo simulation. Users can explore what-if scenarios by adjusting lifestyle sliders (exercise, diet, sleep, smoking, alcohol) and see how those changes relate to their risk education profile, backed by peer-reviewed clinical evidence. This is an educational exploration tool, not a diagnostic or clinical decision aid. It helps users learn about health relationships to have better-informed conversations with their healthcare provider.
Digital Health Twin
Bayesian Causal Network
Simulation
Monte Carlo · 10K
Causal model, not correlation. What-if scenarios grounded in published research.
🧬 Bayesian Inference
Proprietary LLM Infrastructure
Provider abstraction + local AI inference
Multi-provider AI layer enabling seamless switching between cloud (Gemini) and local open-source models (Ollama/MedGemma). 52-case safety evaluation suite with automated benchmarking ensures quality parity across providers. This architecture eliminates single-vendor dependency while enabling on-premise deployment for regulated enterprises that require full data sovereignty.
LLM Infrastructure
Multi-Agent · Fine-Tuned
Active Models
14 LLMs
14 fine-tuned health LLMs with retrieval-augmented generation, consensus safety, and automatic citation.
🧠 Multi-Agent Reasoning
Biosphere World Model
Causal Bayesian network for environmental health risks
A computationally efficient causal Bayesian network with 37+ nodes (fusing 45 peer-reviewed evidence sources and satellite-derived land cover data) that models environment-health risk outcomes like cold exposure, respiratory risk, and infectious disease seasons. Personalised using Conditional Probability Table (CPT) learning from user biometric data and evaluated via a 12-criterion safety consensus engine.
Biosphere World Model
Causal Bayesian Network
Evidence Sources
45 Studies
Causal Bayesian network fusing 45 studies. Models environment-health risk with personalised CPT learning.
🌍 Environment → Health Risk
Satellite Environmental Intelligence
10m-resolution Earth observation for health risk
Integrates Google AlphaEarth Foundations (10m resolution, 64-dim embeddings, CC-BY 4.0), Dynamic World near-real-time land cover classification (9 classes, 2–5 day refresh), and Perch 2.0 on-device bioacoustic species identification (14,795 species, Apache 2.0). Satellite-derived land use, vegetation indices, and impervious surface ratios are fused with ground-level AQI, UV, and meteorological data to produce high-fidelity environmental health risk scores at suburb resolution.
Biosphere World Model
Causal Bayesian Network
Evidence Sources
45 Studies
Causal Bayesian network fusing 45 studies. Models environment-health risk with personalised CPT learning.
🌍 Environment → Health Risk
Built for Production. Backed by Science.
Every capability on this page is live in production, serving real users, and continuously improving. These are not mockups, prototypes, or research experiments. This is infrastructure that works at scale, built by a team that ships. If you're looking for deep tech health AI that's actually deployed, you've found it.
Request a DemoWhy deep tech VCs and enterprise innovation leaders should pay attention.
KuraPath isn't a thin wrapper around a foundation model. It is a vertically integrated health AI platform with eight proprietary deep tech systems that compound into a defensible infrastructure moat. Here is what makes this different.
Defensible IP Portfolio
Nine distinct deep tech capabilities forming a defensible IP portfolio. Following the Aristocrat v Commissioner High Court decision (February 2026), KuraPath's core innovations — including the AI health interpretation methodology, Digital Health Twin, hallucination-mitigated RAG pipeline, and quantum-safe privacy architecture — qualify as patent-eligible under Australian law. On-device health education AI, federated learning with differential privacy, causal health modelling, proprietary LLM infrastructure, and post-quantum cryptography (ML-KEM & ML-DSA) create compounding technical moats that are extremely difficult to replicate.
Production-Grade, Not Research
Everything on this page is live and serving real users today. This is not a research lab demo or a pitch deck concept. These systems are deployed, monitored, and continuously improving in production.
Quantum-Safe Regulatory Compliance
Full compliance with the Privacy Act 1988 (Cth), Australian Privacy Principles, automated decision-making disclosure (Tranche 1 reforms), and SOC 2 security standards. Quantum-safe encryption architecture with AES-256-GCM field-level encryption, crypto-agility versioning (FIPS 197), HMAC-SHA256 pseudonymization (GDPR Art 4(5)), active ML-DSA-65 AI content signing, and deployed ML-KEM-768 hybrid key encapsulation (FIPS 203) and ML-DSA-65 digital signatures (FIPS 204). Data sovereignty in Australian infrastructure.
Compound Competitive Advantage
Each capability reinforces the others. On-device AI enables federated learning. Federated learning improves predictive trajectories. Predictive trajectories feed the digital health twin. Proprietary LLM infrastructure reduces vendor lock-in 60-80%. Quantum-safe encryption protects all data with crypto-agility for future algorithm upgrades. Competitors cannot replicate one without building all eight.
The AI-first vertical healthtech infrastructure layer.
Whether you're an investor evaluating AI-first vertical healthtech opportunities, an enterprise exploring health AI partnerships, or a researcher interested in privacy-preserving health intelligence, we'd love to talk.
Get in Touch