Current maturity
IN DEVELOPMENTTRL 3 engineering proof-of-concept progressing toward TRL 4 integrated laboratory validation.
Medi AI+ Development Programme
Engineering & Project Status
Tracking the progression of Medi AI+ from engineering proof-of-concept toward an integrated laboratory-validated clinical edge AI prototype.
Current Stage · TRL 3 → TRL 4
Pre-clinical engineering
Singapore R&D Programme
Last updated: October 2026
01 / Current status
Current Programme Status
Current release
COMPLETEDv0.7.0-sprint6
Private pre-clinical engineering baseline
Current operating boundary
IN DEVELOPMENTEngineering simulation and controlled development only.
Physical prototype
NOT YET VALIDATEDIntegrated physical Medi AI+ appliance not yet completed.
Clinical status
DEFERREDNot clinically validated.
Regulatory status
NOT YET VALIDATEDNo medical-device regulatory approval claimed.
Hospital deployment
DEFERREDNo production hospital deployment claimed.
02 / Engineering foundation
Engineering Foundation Completed
Medi AI+ already has a substantial software and systems engineering foundation. The current funding and development phase focuses on integrating these capabilities with representative edge hardware, peripherals and validated model candidates.
Platform Core
Completed Engineering Baseline- FastAPI backend architecture
- Persistent application data layer
- Synthetic patient and encounter workflow
- Clinician / supervisor / administrator role controls
- Authentication and session management
- Audit event architecture
- Deterministic simulation provenance
- Containerised runtime
Clinical Simulation
Completed Engineering Baseline- Synthetic patient data only
- Simulated physiological observations
- Representative ECG waveform simulation
- Alert acknowledgement
- Escalation and recovery states
- Supervisor override workflow
- Failure and poor-signal simulation
Speech Workflow
Engineering Candidate- Local/offline speech provider architecture
- 16 kHz mono preprocessing
- Voice-activity metadata
- Immutable unconfirmed ASR draft
- Authenticated human confirmation
- Transcript correction workflow
- Provenance tracking
- Ephemeral audio-processing design
Model adaptation remains gated and has not yet been represented as completed.
Governed Knowledge
Completed Software PoC- Approved-document ingestion
- Immutable knowledge snapshots
- Governed local RAG
- Citation-bound retrieval
- Refusal behaviour
- Source provenance
- Feedback/audit workflow
Safety Architecture
Completed Architecture Foundation- Deterministic warning logic
- Human-review boundary
- Separation of measured facts and generated content
- Fail-closed behaviour
- Provenance-aware context handling
- Governed source hierarchy
Interoperability Foundation
Engineering Baseline- Bounded FHIR R4 compatible JSON structures
- Versioned interoperability contracts
- Encounter export
- Simulated integration boundaries
- Future HL7/HIS adapters defined at architecture level
Current FHIR support is an engineering compatibility baseline and is not represented as certified conformance or live hospital integration.
03 / Architecture status
Frozen Core Architecture
The Medi AI+ core architecture is frozen at product-owner level: a deliberately high-level system boundary that protects the programme’s technical IP while prototype work advances against a stable design.
InputsAuthorized audio, encounter context, device signals and approved knowledge.
Local Medi AI CoreEdge-first processing designed to keep sensitive workflows close to the care setting.
Governed IntelligenceRetrieval, provenance and deterministic safety checks shape every generated context.
Human ReviewClinicians remain the approval boundary for any consequential action.
Reviewable OutputsEvidence-backed documentation and workflow support for qualified users.
IP & governance boundary
The public view stays intentionally abstract. Detailed interfaces, implementation decisions and protected know-how are shared only through a qualified conversation.
The target architecture is human-governed. No AI output independently becomes a diagnosis, prescription, treatment order, confirmed clinical action or final medical record without the appropriate human approval workflow.
04 / Model programme status
AI Model Programme
Medi Speech
Readiness / Adaptation ProgrammePurpose: Local clinical conversational speech processing for edge deployment.
Completed work
- Provider architecture
- Preprocessing pipeline
- Draft/confirmation workflow
- Baseline evaluation framework
- Reproducibility controls
- Governance gates
Current next work
- Approved dataset admission
- Rights and consent verification
- TRAIN / DEV / SEALED TEST separation
- Medical-domain adaptation
- Candidate evaluation
- Edge conversion
- Jetson validation
Project-specific model-weight adaptation has not yet been represented as completed.
Medi Knowledge
Governed Knowledge DevelopmentPurpose: Evidence-grounded local clinical knowledge resolution.
Completed work
- Governed RAG baseline
- Source versioning
- Citations
- Refusal
- Provenance controls
- Knowledge snapshot architecture
Current next work
- Institutional source admission
- Medical Knowledge Graph construction
- Retrieval/reranking evaluation
- Governed local generation candidate
- Edge optimisation
- Institutional validation
Knowledge Graph construction is not the same as model training.
05 / Physical prototype status
Physical Prototype Programme
The next major programme milestone is integration of the existing software engineering foundation with representative edge compute and medical/peripheral interfaces.
IN DEVELOPMENT / FUNDING STAGEEdge Compute
Jetson Orin-class platform
Hardware validation pending
User Interface
Integrated touch display / clinical workstation interface
Software UI foundation available; final appliance integration pending
Audio
Microphone / care-zone audio input
Software processing architecture available; physical deployment validation pending
Physiological Inputs
ECG, SpO₂, heart rate, temperature and other approved peripherals
Representative physical integration pending
Device Interface
STM32 / governed sensor gateway
Simulated protocol foundation completed; physical HIL validation pending
Enclosure
Industrial/prototype enclosure
Development pending
06 / TRL progression
TRL 3 → TRL 4 Development Path
Current Foundation
Engineering proof-of-concept
- Software subsystems
- Simulation workflows
- Architecture contracts
- Safety/governance foundation
- Edge packaging
- Synthetic scenario testing
Integration and laboratory verification
Funding and development work
- Hardware integration
- Representative sensors
- Model adaptation
- KG development
- Edge optimisation
- Interoperability simulation
- Performance testing
- Safety/failure verification
Integrated laboratory prototype
Target evidence
- Software + hardware operating together
- Representative sensors/peripherals integrated
- Core AI workflow on target-class edge platform
- Controlled synthetic/engineering scenarios
- Measured latency and resource usage
- Thermal/power observations
- Failure/recovery validation
- Reproducible prototype configuration
TRL 4 does not equal clinical validation or regulatory approval.
07 / Programme tracker
Programme Milestones
Phase 1
Core Engineering Foundation
COMPLETEDSprints 0–6 · Simulation platform · Governed RAG · Speech engineering workflow · Safety architecture · Audit and data governance foundation
Phase 2
Architecture Alignment
COMPLETEDFrozen Core architecture · ClinicalContextEvent contracts · Source-authority model · HIS/FHIR boundary · Human approval boundary
Phase 3
Model Readiness
IN PROGRESSDataset governance · Adaptation readiness · Evaluation protocol · Model candidate preparation
Phase 4
Physical Integration
FUNDING / DEVELOPMENT TARGETJetson · Display · Audio · Medical/peripheral inputs · Sensor gateway · Enclosure
Phase 5
Integrated Laboratory Prototype
TARGETComplete TRL 4 integration · Laboratory verification · Edge performance evaluation
Phase 6
Institutional Validation
FUTURE STAGEHospital/research collaboration · Interoperability evaluation · Clinical workflow validation · Regulatory assessment
Phase 7
Controlled Deployment
FUTURE / SUBJECT TO APPROVAL08 / Engineering evidence
Current Engineering Evidence
Engineering release
v0.7.0-sprint6
Software programme
Sprints 0–6 engineering baseline completed.
Regression evidence
394 backend tests reported at the latest architecture-alignment gate.
Packaging
AMD64 and ARM64 container support.
RAG
Governed local approved-document RAG demonstrated in engineering simulation.
Speech
Offline speech engineering workflow demonstrated with human confirmation.
Safety
Deterministic safety/warning architecture with audit provenance.
Simulation
Synthetic patient and device-data workflows.
Recovery
Engineering backup/recovery and retention controls.
Engineering test counts are software verification evidence only and must not be interpreted as clinical performance metrics.
09 / Claim boundary
Explicitly Not Yet Claimed
Medi AI+ does not currently claim:
Clinical validation
Diagnostic accuracy
Treatment recommendation capability
Autonomous clinical decision-making
Regulatory approval
Production hospital deployment
Certified FHIR conformance
Production HIS write-back
Completed Jetson performance validation
Validated physical medical-device interfaces
Completed project-specific model adaptation
Commercial readiness of the final appliance
These items belong to later engineering, institutional and regulatory stages.
10 / Current funding objective
S$200k Prototype & Validation Round
Advance Medi AI+ from TRL 3 engineering proof-of-concept toward TRL 4 integrated laboratory validation.
Hardware Integration
- Edge compute
- Display
- Audio
- Representative peripherals
- Interface electronics
AI Adaptation
- Medi Speech
- Medi Knowledge
- Evaluation
- Edge conversion
Systems Integration
- Core Engine
- Device ingestion
- Interoperability
- Knowledge integration
Verification
- Latency
- Resource usage
- Power/thermal
- Recovery
- Safety
- Reproducibility
Target outcome: An integrated laboratory prototype ready for the next stage of institutional evaluation.
11 / Partnership needs
Current Partnership Priorities
Clinical / Research Institutions
- Workflow validation
- Use-case definition
- Future institutional evaluation
Medical Technology Partners
- Sensors
- Peripherals
- Device interfaces
AI / Edge Infrastructure Partners
- Edge optimisation
- Deployment tooling
- Inference acceleration
Regulatory / Quality Partners
- Intended-use definition
- Classification
- Quality planning
- Cybersecurity pathway
Investors & Innovation Programmes
- TRL 3 → 4 prototype funding
- Technical validation
- Commercialisation support
12 / Version and status
Technical Status Index
A compact readout of the current engineering boundary: what is established in software, what remains in development, and what is intentionally deferred until later validation stages.
01
Core Platform
Engineering baseline complete
02
Synthetic Simulation
Complete
03
Authentication / RBAC
Complete
04
Audit / Provenance
Complete
05
Deterministic Alerts
Complete
06
Governed RAG
Engineering PoC complete
07
Offline Speech
Engineering candidate
08
Medical KG
In development
09
Model Adaptation
Readiness stage
10
Jetson Deployment
Pending physical validation
11
Physical Sensors
Pending integration
12
Hospital FHIR
Not yet implemented
13
Production HIS
Not yet implemented
14
Clinical Validation
Future stage
15
Regulatory Approval
Future stage
13 / Safety boundary
Current Safety Boundary
Medi AI+ is currently a pre-clinical engineering programme operating with synthetic and controlled engineering inputs. The current system is not authorised for diagnosis, treatment, prescribing, autonomous clinical decision-making or unsupervised patient care.
Any future handling of live patient data, hospital integration or clinical use will require separate technical, institutional, privacy, cybersecurity and regulatory approval.
14 / Continue reading