What This Page Is
A concise map of how Ukubona turns evidence into tools: Inputs → Model → API → UI. We ship Flask apps with backend APIs that connect to public datasets and, when appropriate, private data under strict disclosure-risk controls. The output isn’t a decree — it’s a place to rehearse decisions.
Pipeline
Inputs
Registries, trials, vetted literature, context. Curated and versioned.
Model → API
Validated equations (e.g., survival/Cox) served via Flask endpoints. Assumptions visible; uncertainty preserved.
UI (Rehearsal)
Interactive loops (avatars, sliders, comparators) to play → learn → adjust → decide.
Ukuvula — Provisioning
Social Determinants Ingestor
SDOH vectors & deprivation indices for a place.
Nutrition Exposure
Access/affordability buckets + elasticity.
Env. Normalizer
Air/heat/traffic indices normalized for downstream risk models.
Ukuzula — Mobility & Vitality
Vitality Signal
Steps/cadence → vitality score & slope.
Gait → Risk Mapper
Simple mobility signals to absolute-risk proxies with uncertainty.
Recovery Curve
Post-event recovery estimator from short time-series windows.
Ukukona — Play & Experiments
Scenario Engine (Avatars)
Branch outcomes, tails, replays — safe rehearsal before deployment.
Behavior Loop
Nudges → adherence curves; optimal micro-commitments.
Cost of Delay
Time-to-action vs outcome curves to expose hidden tail risks.
Ukubona — Integration/Nervous System
Comparator & Benchmarks
Your estimate vs matched peers; explainable deltas.
Consent-as-Rehearsal
Snapshot of what was practiced & understood; versioned, auditable.
Audit Digest
Human-readable audit reports: inputs, transforms, outputs, return.
Ukukula — Aging & Legacy
ADL Trajectory
12–24 mo ADL bands; care-plan triggers.
Family Impact
Caregiver time/$ projections; decision surfaces.
Preference Elicitor
Lightweight survey loop to align plans with values.
Cross-Cut — Energy & Simulation Cost
Energy Cost Monitor
kWh/$ per clarity; marginal cost curves as simulations scale. Clarity has a price.
Compute Budget Planner
Plan runs vs budget; surface trade-offs before launch.
Footprint Reporter
Per-run & cumulative footprint summaries for stakeholders.
Data Access & Disclosure Risk
Public Data
Documented endpoints; no identifiers; aggregation/suppression when needed.
Private Data
Access-controlled. Min-fields, hashed/bucketed quasi-IDs, row-level guards.
Disclosure Controls
k-anon thresholds, top/bottom-coding, small-cell noise, rare-combo redaction.
API — Illustrative Shape
Example: request an estimate + comparator band from a Flask endpoint.
// POST /api/v1/risk/estimate
{
"model": "periop_90d",
"inputs": { "age": 40, "sex": "F", "race": "White", "health": "very_good" },
"comparators": ["matched_nondonor"]
}
// 200 OK
{
"estimate": { "risk_90d_per_10k": 2.2, "ci": [1.0, 4.7] },
"comparator": { "matched_nondonor_per_10k": 1.3, "delta": 0.9 },
"explain": { "equation": "cox", "assumptions": ["PH ok"], "version": "2025.08" }
}Private variants add auth headers and stricter logging; public endpoints return only safe aggregates.
Cost of Clarity
More simulation → more intelligence → more decisions — and more energy and money. We surface these trade-offs so teams invest where value is highest and avoid “exhaust” where it isn’t.
Build With Us
Want your model behind an API and into a rehearsal UI? We’ll deploy a Flask app, wire disclosure controls, and ship a clean interface.