Engineering Plan
24 epics, ranked P0–P3 by Atlas-parity priority, mapped to a 7-sprint timeline. Yogini stops using Atlas at month 7.
Priority bands
The 24 epics from the launch briefing, re-ranked by Deepak in the polish round so the team grooms in priority order.
P0 — Yogini-can't-switch-without (11 epics)
Foundation work. Without these, Yogini cannot drop Atlas. Everything in this band is table-stakes parity; the wedge moves come in P1.
| Epic | What |
|---|---|
| Foundation | Repo, CI, deploy pipelines |
| Auth | Cognito user/practice flows, session, role gates |
| Chart shell | Patient header + 13 left-rail tabs (Summary, Timeline, Notes, Messages, Tasks, Appointments, Documents, Diagnoses, Prescriptions, Labs, Vitals & growth, Bright Futures, Billing) |
| SOAP editor | Inline edit, ICD-10 chips, sign-and-add |
| Vitals | Pediatric growth charts, weight/height/BMI tracking |
| eRx | DoseSpot integration, refill queue, controlled-substance gating |
| Calendar | Week/day view, color-coded visit types, prospect calls |
| Stripe billing | Subscriptions, family rollups, MRR/ARR |
| SMS + email | Twilio for SMS, transactional email, audit trail |
| Staff-add-patient | Front-desk intake flow |
| Documents | Upload, tag, attach to chart, e-sign HIPAA + DPC |
P1 — Wedge vs Atlas (5 epics)
The reasons to switch. P0 makes parity possible; P1 makes Atlas indefensible.
| Epic | What |
|---|---|
| Prospect onboarding | Magic-link → family registration → e-sign → Stripe Setup Intent → first-visit booking |
| Parent app + auto-summary | The leapfrog. Push notification with plain-language visit summary auto-generated by Claude |
| AI scribe | Whisper transcription + Claude SOAP draft + confidence flags |
| AI triage + risk | Inbox prioritization, drug-interaction checks, draft replies |
| Newborn home visit | 48h post-discharge premium flow with structured measurements + recording |
P2 — Modern polish (3 epics)
Quality-of-life that compounds with the wedge.
P3 — Expansion (5 epics)
Future surface area: new specialties, multi-provider practices, billing analytics deepening, etc.
Sprint mapping
7 sprints (sprint-0 through sprint-7+). Targeting Yogini stops using Atlas at month 7.
| Sprint | Focus |
|---|---|
| sprint-0 | Foundation, auth, chart shell scaffold |
| sprint-1 | SOAP editor, vitals, calendar |
| sprint-2 | Documents, Stripe billing |
| sprint-3 | eRx (DoseSpot), SMS + email, staff-add-patient |
| sprint-4 | Prospect onboarding, parent app shell |
| sprint-5 | AI scribe (Whisper + Claude SOAP draft) |
| sprint-6 | Auto-summary push, AI triage + risk |
| sprint-7+ | Newborn home visit, polish, expansion |
The mapping is in the GitHub epic labels; this table is the human-readable summary.
Erik's read on the timeline
"Most likely the epics you have identified will go FSTR than you have projected — because they are well defined and we have clickable screens."
The clickable prototype removes most ambiguity. Each epic ships against a known target UI.
"We need to explicitly create engineering time for Security and Regulation Compliance."
P0–P3 doesn't yet line-item HIPAA pilot work, BAA negotiations with subprocessors, audit prep, or pen-test cycles. We need fractional bandwidth from Poy/Allan for security and infrastructure (covered in Decisions, item 2) and at least one fractional QAA — ideally two.
GitHub backlog
Live in dsurana04/starlight-md.
Supporting docs (in the launch page)
The launch briefing prototype includes a bottom-section docs library:
| Doc | Purpose |
|---|---|
| MVP definition | What ships first |
| Atlas teardown | Where Atlas falls short, evidence-backed |
| v1 priorities | The wedge moves |
| Record / documents audit | Drizzle schema for clinical records |
| Visit flow spec | The end-to-end visit interaction model |
| 18-patient roster | Sample data for the prototype |
| Chart design rationale | Why the 13-tab left rail looks the way it does |
| 4 QA agent reports | Findings from the parallel QA pass |
| Epic backlog | The 24 epics, expanded |