Six phases from foundation to NHI capitation at scale. Phases 0–2 engineering is substantially complete (~R27.1M invested). 12–18 months to first revenue. 48–54 months to EBITDA breakeven. A structured, milestone-driven path to national healthcare impact.
The SAHPRA clinical validation sequence is the critical path. Everything else — scheme partnerships, physician onboarding, infrastructure build — runs in parallel and feeds into this gate.
LLM selection & fine-tuning on SA corpus
Retrospective validation study
Prospective pilot — 3 sites, ≥1,500 consultations
SAHPRA file complete → First revenue
Quality gates for SAHPRA: AI sensitivity ≥92% on red-flag conditions, specificity ≥88% on EML-compliant prescribing, physician override rate <15% across all disease categories, bias parity within 5% across demographic groups.
The team grows in step with each phase's requirements. Phase 0's 10-person core expands to ~35 by Phase 4 as engineering, physician network, and commercial functions scale.
| Phase | Total Headcount | Engineers | Physicians (Network) |
|---|---|---|---|
| Phase 0 (Month 0–3) | 10 | 6 | 0 (in hiring) |
| Phase 1 (Month 3–12) | 14 | 8 | 5–10 (pilot) |
| Phase 2 (Month 12–18) | 18 | 10 | 30+ |
| Phase 3 (Month 18–30) | 24 | 14 | 50–100 |
| Phase 4 (Month 30–54) | 35 | 20 | 200+ |
Team scaling from current position: Bridge round funds leadership expansion (Regulatory Lead, BD Lead, Clinical Lead) and initial physician network recruitment. Team grows from current technical core.
The technical build runs across seven workstreams in Phase 1, all converging at the Phase 2 prospective pilot gate.
FHIR R4 data store, consent ledger, audit log, SA ID integration, POPIA-compliant data residency on GCP africa-south1 (Johannesburg).
LLM fine-tuning on SA corpus, deterministic rules engine (EML/ICD-10/STG), red-flag escalation logic, explainability chains.
RabbitMQ case queue, physician review dashboard, approve/modify/escalate UI, HPCSA verification integration, audit trail.
React 19+ PWA, audio mode, multilingual clinical flow (11 languages).
Healthbridge claims API, pharmacy e-prescription network (MediFin/ScriptWise), NHI readiness FHIR API.
MLflow experiment tracking, vLLM/TGI serving, continuous learning pipeline, physician override signal processing, monthly bias audits.
Kubernetes (GKE), Terraform IaC, GitHub Actions CI/CD, Spring Boot Actuator monitoring, penetration testing, ISO 27001 preparation.
Each phase has defined exit criteria. Funding tranches and investor reporting are tied to these gates.
HPCSA, SAHPRA, POPIA, and NHI compliance are designed into Aafiya's architecture — not bolted on. Explore how Aafiya meets every regulatory requirement.