ColdMesh 1.0 serves quantile dwell, travel, and locker-diversion models that feed StochVRP-Mesh routing — ischaemic budgets, VRPTW windows, and quantum-ready pathfinding on Kryptur infrastructure.
Kryptur advances smarter medical logistics through quantile risk estimates, cold-chain countdown synthesis, and quantum-ready routing under hard ischaemic deadlines.
Live prediction console for dwell, travel, and locker diversion with cold-chain risk read-out.
22-feature LightGBM quantile regressor — 80% PI coverage 0.747 for facility handover time.
12-feature travel-time model trained on 80,000 samples with 80% PI coverage 0.771.
Binary success classifier at ROC-AUC 0.849 against a 52% base success rate.
Authenticated JSON endpoints for dwell_advanced, travel_time, and locker_advanced.
Predictive layer feeding classical VRPTW and QAOA pathfinding for organ delivery.
Advanced dwell quantile model covers facility handover risk under ischaemic remaining time.
Travel-time quantiles absorb congestion, weather, and peak-hour effects across dense corridors.
Diversion success probability guides locker handoff when route slack and occupancy collide.
Classical multi-vehicle routing with hard ischaemic windows across 15 stops and 4 vehicles.
Same-origin /api/coldmesh forwards authenticated inference to the StochVRP-Mesh backend.
Optimise cold-chain outcomes from quantile prediction and NP-hard routing investments.
Same-origin proxy at /api/coldmesh forwards JSON payloads to the StochVRP-Mesh prediction backend. Endpoints: /predict/dwell_advanced (22 features + cyclic time encodings), /predict/travel_time, and /predict/locker_advanced. Responses return quantile minutes (q10/q50/q90) or diversion success probability for cold-chain risk synthesis.
Quantile‑regression risk estimates for dwell time, travel time and locker diversion — the predictive layer of a quantum‑ready operating system for time‑critical medical cargo.
Classical VRPTW solver returning Pareto‑optimal routes across mileage, time‑window adherence and driver‑workload equity.
10,000‑replication Monte Carlo engine reporting conditional value‑at‑risk of tardiness from a vine‑copula travel model.
QUBO formulation for >2,000‑stop instances, solved via VQE on cloud QPUs with classical 2‑opt polishing as fallback.
What's New at Kryptur — ColdMesh 1.0 API releases, organ logistics research, and VRPTW news.