Working prototype
Renal medicine
Dialysis CoPilot
Surface the modifiable physiology behind mortality and admission risk across a dialysis unit.
- Vascular access failure
- Fluid overload mitigation
- Intradialytic hypotension
Clinical decision support · Research prototype
Models a clinician
can argue with.
Decision support that shows you what to change, not just what will happen. Counterfactual clinical AI for high-stakes nephrology, neurovascular, and interventional decisions.
Causal inferencePhysiology-groundedInteractive counterfactuals
01
Working module
09
Dialysis pathways
03
Modules in pipeline
Interactive prototype
Interrogate the intervention.
Constructed patient data for demonstration only. Outputs are simulated projections, not clinical recommendations.
Dialysis counterfactual workspace
Working prototype · In-centre haemodialysis
Clinical patient cardSynthetic
KUV-HD-0482
64y · Type 2 diabetes · 4.2y on HD
CVC access
Hb 9.8 g/dL
Albumin 3.2 g/dL
hs-CRP ↑
Primary risk signal
Mortality, admission and haemodynamic stability
Counterfactual interventions
Dry weight reduction−1.4L
−1.0L−2.0L
Change any assumption. The projected outcomes and mediated pathway update together.
Projected impact
30-day mortality
10.9%↓ 1.9 pp
0Baseline 12.8%
90-day mortality
18.3%↓ 3.1 pp
0Baseline 21.4%
1-year mortality
34.3%↓ 4.3 pp
0Baseline 38.6%
30-day hospitalisation
27.2%↓ 4.0 pp
0Baseline 31.2%
Patient risk trajectory
Baseline mortality, counterfactual mortality and admission risk over time
Physiological pathway cascade
Mediated effects under the selected counterfactual
Systemic inflammation
Endothelial shear
Myocardial stunning
Cardiac event risk
The premise
“A risk score that cannot be acted on is just anxiety with a number attached.”
Traditional predictive ML
Correlation dressed as instruction
The confounding trap
The sickest get the strongest treatment. A model may learn that treatment predicts harm, not that it prevents it.
Feedback loops
Outcomes reflect old practice. When care changes, improvement can look like model failure.
A score identifies risk but leaves the clinical lever hidden.
Kuviyam causal AI
Mechanism made inspectable
01
Read the trajectory
Labs, sessions, medications and events over time—not one convenient snapshot.
02
Route through physiology
Explicit mechanisms with named drivers, thresholds and uncertainty.
03
Test the change
Rank interventions by the risk they may remove for this patient.
Dialysis first · Specialty pipeline
Deep in dialysis. Building carefully beyond it.
Dialysis CoPilot is the active working prototype. Stroke, liver and renal cancer modules remain in the clinical development pipeline.
In pipeline
Neurovascular
Stroke Pathway CoPilot
Make time, tissue and haemorrhage trade-offs visible while the treatment window is still open.
- Penumbra salvage
- Haemorrhagic transformation
- Thrombectomy windows
Clinical development pipeline
In pipeline
Interventional radiology
Liver Intervention CoPilot
Test HCC ablation and embolisation choices against functional liver reserve before treatment.
- HCC ablation planning
- Decompensation risk
- Functional reserve
Clinical development pipeline
In pipeline
Interventional oncology
Renal Cancer CoPilot
Balance biopsy, ablation margin and nephron protection for small renal masses.
- T1 renal mass
- Collecting-system safety
- Kidney function
Clinical development pipeline
Evidence & governance
Prototype today.
Registry next.
Nothing claimed before it is earned.
The pathway map is a clinical document: reviewable, versioned and open to challenge. Every number shown here comes from constructed test cases—not real patients.
Observational benchmarking
Test against defined cohorts, agreed endpoints and known standards before deployment.
Prospective registry
Track predictions, interventions and outcomes forward—not only against historical records.
IRB-ready governance
Clear data lineage, protocol versioning, access control and institutional review readiness.
Clinician sovereignty
Decision support, never prescription. Uncertainty and missingness remain visible.
Institutional pilot
Bring your hardest cases.
Challenge the model.
We are looking for clinical directors, nephrology and radiology unit heads, and research partners who want assumptions on the table—not buried in a black box.
hello@kuviyam.ai