See the liver. Predict the future. Intervene earlier — now on Oracle Cloud.
LiverTwin™ fuses multi-omics, ultrasound/MRI, and longitudinal EHR to generate a patient-specific digital twin—optimized for price/performance and privacy on Oracle Cloud Infrastructure (OCI).
The Challenge
MASLD/MASH affects millions but remains critically underdiagnosed, leading to preventable complications and costs.
Up to 80% of MASLD/MASH cases go undetected until advanced stages, missing critical intervention windows.
Clinicians juggle disconnected systems—labs, imaging, EHRs—without unified patient insights or predictive guidance.
By the time symptoms appear, treatment costs skyrocket while patient outcomes decline dramatically.
What LiverTwin™ Does
Transform fragmented patient data into actionable insights with AI-powered digital twins.
Fuse multi-omics, imaging, and EHR data into comprehensive patient-specific models.
Predict disease trajectory and intervention timing with validated AI algorithms.
Receive evidence-based recommendations with clear clinical rationale and confidence scores.
Seamlessly integrate insights into existing clinical workflows without disruption.
Why LiverTwin™ Is Different
Built from the ground up for clinical reality, powered by Oracle Cloud's enterprise-grade infrastructure.
Multimodal by Design
Native multi-omics fusionUnlike single-modality solutions, LiverTwin™ natively integrates genomics, proteomics, imaging, and clinical data for comprehensive patient understanding.
GPU-Accelerated Foundation
Real-time AI inferenceLeverages Oracle's H100/H200/GB200 Supercluster for real-time inference and continuous model updates, ensuring predictions stay current with latest research.
Explainable Outputs
Clinical-grade explainabilityEvery prediction comes with clear clinical rationale, confidence intervals, and supporting evidence—building trust through transparency.
Core Modules
Six specialized AI modules work together to create comprehensive digital liver twins.
Process genomics, proteomics, and metabolomics data into unified representations for disease risk assessment.
Analyze ultrasound, MRI, and CT scans to quantify liver structure, fat content, and fibrosis progression.
Transform longitudinal EHR data into temporal patterns revealing disease progression dynamics.
Intelligently combine multi-modal inputs using attention mechanisms for holistic patient understanding.
Run forward simulations to predict disease trajectories under different intervention scenarios.
Generate actionable recommendations with clinical rationale tailored to individual patient contexts.
Reference Architecture on Oracle Cloud
Enterprise-grade infrastructure designed for healthcare AI workloads with PHI-by-default privacy and performance optimization.
Hub-and-Spoke VCN Design
LiverTwin™ leverages Oracle Cloud's hub-and-spoke Virtual Cloud Network (VCN) architecture to ensure secure, scalable, and compliant healthcare AI operations.
All patient health information (PHI) remains within private networks by default, with dedicated private endpoints ensuring no data exposure to public internet.
Private Networking
FastConnect and IPSec VPN provide deterministic, private connectivity with healthcare providers and research institutions.
Specialized Spokes
Dedicated network segments for data platform, AI/ML training, inference applications, and imaging workflows.
Security by Design
Multi-layered security with WAF, Network Firewall, Security Zones, and comprehensive audit logging.
Performance & Cost Levers
Mixed Precision
FP16/BF16 training reduces memory usage and accelerates training by 2x
RDMA Fabric
Ultra-low latency networking for distributed training across GPU clusters
Local NVMe
Large local storage for checkpoints eliminates network bottlenecks
Autoscaled Pools
Dynamic GPU node scaling based on workload demands
Data Locality
Co-location within same availability domain reduces latency
Service-by-Service Mapping
Comprehensive mapping of LiverTwin™ functions to Oracle Cloud Infrastructure services with alternatives and implementation notes.
| Distributed training | OCI Supercluster (H100/H200/Blackwell GB200) | NVIDIA AI Enterprise; DGX Cloud on OCI as fallback. |
| Checkpointing/local NVMe | BM.GPU.* local NVMe | Durable replicas on Block Volumes. |
| Experiment tracking & jobs | OCI Data Science (Jobs, Pipelines, Model Catalog) | MLflow on OKE (self-managed). |
| Inference (real-time/batch) | Data Science Model Deployments (Private Endpoints) | OKE + Triton/KServe; NVIDIA NIM tutorial pattern. |
| RAG/Vector DB | Oracle Database 23ai – AI Vector Search | MySQL HeatWave Vector; OpenSearch vector. |
| Data lake & ETL | Object Storage, Data Flow (Spark), Data Integration, GoldenGate, Streaming | Service Connector Hub for events. |
| Imaging bridge | OKE + Orthanc/dcm4chee; File Storage/Object Storage | DICOM SR write-back; FHIR notes. |
| API layer | API Gateway, Functions | APEX for ops apps. |
| Observability | Logging, Monitoring/Alarms, OpenSearch | APM optional. |
| Security & keys | IAM, Vault/KMS (HSM), Cloud Guard, Security Zones, WAF, Network Firewall, Data Safe | Bastion, Private Endpoints. |
| Multicloud models | OCI Generative AI (Cohere, Meta, Mistral, Google Gemini) | BYO models on OKE/Data Science. |
Data Science deployments accessible only inside VCN, ensuring patient health information never traverses public networks.
Private access to Object Storage and OCI services without public egress, maintaining complete network isolation.
Deterministic, high-bandwidth connectivity with healthcare providers and research institutions via dedicated circuits.
Migration to OCI
Systematic six-phase approach to migrating LiverTwin™ from Lightning.ai to Oracle Cloud Infrastructure with minimal disruption and maximum performance optimization.
Compartments, IAM, Cloud Guard + Security Zones; hub-and-spoke VCN; Bastion; Service Gateway; WAF/Network Firewall; OCIR; Data Science project; small OKE cluster.
Exit Criteria
Posture green; private paths tested.
Performance & Cost Levers
Lightning.ai → OCI (Model & Pipeline Translation)
Comprehensive mapping of Lightning.ai components to Oracle Cloud Infrastructure services for seamless migration and enhanced performance.
| Current (Lightning.ai) | Target on OCI | Notes | |
|---|---|---|---|
Lightning Trainer Training | OCI Supercluster (H100/H200/Blackwell) via Data Science Jobs/Slurm | Map NCCL/FSDP/DeepSpeed to RDMA fabric | |
Checkpoints Storage | Object Storage (+ lifecycle to Archive) | Private access via Service Gateway | |
Experiment tracking MLOps | Data Science Model Catalog (or MLflow on OKE) | Metrics + lineage | |
Batch inference Inference | Data Science Jobs | Private networking to DB/vector store | |
Real-time inference Inference | Data Science Model Deployments (Private Endpoints) or OKE + Triton/KServe | PHI stays inside VCN | |
RAG store Data | Oracle Database 23ai – AI Vector Search | SQL + vector with metadata filters | |
Streaming/queues Integration | OCI Streaming (Kafka-compatible) | Event-driven architecture | |
CDC pipelines Data | OCI GoldenGate to Object Storage/ADB | Real-time data synchronization | |
Imaging Healthcare | Orthanc/dcm4chee on OKE; DICOM SR back to PACS | Healthcare imaging workflow |
Categories
Data Movement to OCI
Systematic approach to migrating healthcare data to Oracle Cloud Infrastructure with enterprise-grade security and AI optimization.
PHI vs non-PHI; hot vs cold
Categorize data by sensitivity and access patterns to optimize storage and compliance strategies.
Data Transfer Service/Roving Edge → Object Storage (raw)
Migrate large historical datasets using Oracle's secure transfer services for efficient bulk loading.
FastConnect/IPSec + GoldenGate CDC + Streaming
Establish real-time data pipelines for continuous synchronization and change data capture.
Spark → Data Catalog; generate embeddings → DB 23ai
Process and enrich data using Spark, catalog assets, and generate AI-ready embeddings.
Data Flow Architecture
Secure Transfer
PHI remains encrypted and private throughout the entire migration process
Minimal Downtime
Bulk historical transfer followed by real-time CDC ensures continuous operations
AI-Optimized
Data automatically processed into vector embeddings ready for ML workloads
Security & Governance
Enterprise-grade security controls and compliance frameworks designed for healthcare AI workloads.
Complete infrastructure compliance with healthcare privacy regulations and audit requirements.
Role-based access controls ensuring users only access necessary resources and data.
Customer-managed encryption keys with hardware security module backing for maximum control.
Continuous security posture monitoring with automated threat detection and remediation.
Preventive guardrails that automatically block non-compliant configurations and deployments.
Multi-layered network protection with application-level filtering and DDoS mitigation.
Database security assessment, user activity monitoring, and sensitive data discovery.
Tamper-proof audit trails with immutable logging for compliance and forensic analysis.
Automated de-identification workflows for DICOM/FHIR data to protect patient privacy.
Traffic Stays Private by Default
All patient data flows through private networks via VCN + Private Endpoints + Service Gateway, ensuring zero exposure to public internet.
Oracle Ask
Engage with Oracle's AI experts to get personalized guidance and support for your LiverTwin™ deployment.
Get expert advice on AI model optimization and deployment strategies.
Conduct a thorough security audit to ensure compliance with healthcare regulations.
Review the explainability of your AI predictions to build trust with stakeholders.
Analyze your cloud costs and optimize your budget for LiverTwin™.
Evaluate your Oracle Cloud Infrastructure setup for optimal performance and security.
Provide training and support for your users to effectively utilize LiverTwin™.
Free Tier & Uplift
Start with Oracle Cloud's free tier and scale up as needed for your LiverTwin™ deployment.
Access to essential Oracle Cloud services at no cost to evaluate LiverTwin™.
Easily scale your infrastructure to meet growing demands.
Get support for your free tier and any uplifts needed for LiverTwin™.
Train your AI models using Oracle Cloud's powerful GPU resources.
Benefit from Oracle Cloud's robust security features to protect your data.
Use Oracle's explainability tools to understand your AI predictions.
Cutover Checklist
Ensure a smooth transition to Oracle Cloud with our detailed cutover checklist.
Backup all your existing data before migration.
Test your AI models in the Oracle Cloud environment.
Configure Oracle Cloud's security features to meet your requirements.
Estimate your cloud costs and plan your budget accordingly.
Set up your Oracle Cloud infrastructure for LiverTwin™.
Train your users on the new Oracle Cloud environment.
Integrations & Deployment
Seamless integration with existing healthcare systems and flexible deployment options for diverse organizational needs.
EHR Standards
Imaging Protocols
Cloud Deployment
AI Models
Container Registry
OCIR for secure container image storage and distribution
Infrastructure as Code
Terraform modules for reproducible deployments
CI/CD Pipeline
GitHub Actions with OCI DevOps integration
Measurable Impact
Clinical validation demonstrates significant improvements in detection rates, care efficiency, and patient outcomes.
Increase in early-stage MASLD/MASH identification compared to standard care protocols
Reduction in time from initial assessment to specialist referral and intervention
Improvement in patient compliance with lifestyle and therapeutic interventions
Per 1,000 patients through early intervention and prevention of advanced disease
Frequently Asked Questions
Common questions about LiverTwin™ implementation, deployment, and clinical integration.