Solara · Platform

A sales network that compounds on its own output.

Three systems, one loop. Territory intelligence, rep optimization, and adoption analytics feed each other in a closed cycle — every rep conversation in the field is training data for the next one. The floor rises every quarter.

The Karpathy Loop

Every rep interaction is training data.

Traditional rep networks treat each call as an isolated event. Solara treats every call, visit, sample, and outcome as a single unit of training data for a model that runs across the whole network. When a diabetes educator in Tampa books a conversion, the model updates targeting and scripts for the rep walking into a similar practice in Phoenix next Tuesday.

The loop has four steps and runs continuously. It's the reason the system compounds.

/01 Interaction
Rep calls physician
/02 Outcome
System logs result
/03 Update
Model retrains
/04 Dispatch
Network re-targets
The
loop
Pillar / 01

Territory Intelligence— I.

Solara builds a live provider universe for every category in the network, then filters it against each manufacturer's clinical indications and commercial priorities. Territories are drafted against real opportunity density, not legacy map-shape politics — and they rebalance as the data moves.

  • Data
    Unified provider graph.

    CMS NPPES, Part D + Part B claims, specialty taxonomy, affiliation + practice data — resolved to a single scorable provider entity across 9.3M records indexed through the Thunder Holdings data platform.

  • Matching
    Indication-to-prescriber fit.

    Manufacturers describe their target indication; the model ranks providers by likely fit based on prescribing pattern, diagnosis mix, and payer alignment.

  • Geography
    Opportunity-balanced territories.

    Drawn from opportunity density and drive-time reality, not state lines. Rebalance triggers fire when coverage drifts beyond tolerance.

  • Refresh
    Weekly data refresh.

    Provider state updates weekly. Big moves (affiliation changes, new DEA status) trigger reviews out-of-cycle.

Pillar / 02

Rep Optimization— II.

This is where the loop lives. Rep outcomes feed a model; the model pushes tighter targeting, refreshed scripts, and next-best-action prompts back to every rep in the network. The best reps' intuition stops being trapped in their heads — it propagates.

  • Ranking
    Ranked call lists, per rep.

    Every morning the platform surfaces the 20 highest-probability conversations in the rep's territory, ordered by adoption likelihood × product-fit score.

  • Scripts
    Category-specific playbooks.

    Clinical talking points, objection responses, and evidence packets per indication. Updates flow when the network's response rate shifts.

  • Prompts
    Next-best-action between visits.

    When a conversation stalls, the platform suggests the follow-up sequence that has unstuck similar stalls elsewhere in the network.

  • Training
    Outcomes become training data.

    Every adopt / stall / no-go annotation feeds the model. The network's floor rises every quarter — and rep rankings surface the operators driving it.

Pillar / 03

Adoption Analytics— III.

Manufacturer executives see physician-level adoption state, stall and churn risk scoring, and rep contribution attribution in real time — not stitched together from monthly exports. The same telemetry exports to your data warehouse on a schedule you control.

  • State
    Physician adoption state, live.

    Provider-level view: evaluating, adopting, stalled, churned, won-back. Filter by territory, rep, or specialty.

  • Risk
    Stall + churn scoring.

    Model-flagged accounts where the next interaction matters disproportionately. Triaged to the right rep with a recommended play.

  • Attribution
    Rep contribution attribution.

    Clean contribution accounting per rep per account. Incentive alignment stops being a forensic exercise.

  • Export
    Warehouse export.

    Scheduled pushes to Snowflake, BigQuery, or Redshift. Event-level detail, not dashboard screenshots. Your data stays yours.

Under the Hood

The stack, documented.

Solara is transparent about how the platform is built. Your IT and security teams won't be decoding a black box — the foundation is boring enterprise infrastructure, by design.

/01 · Application
Vite + React 19 · Fastify · TypeScript
  • ·Vite + React 19 marketing + rep portal
  • ·Fastify API on Railway
  • ·Type-safe end-to-end
/02 · Data
Postgres · Thunder data platform
  • ·Managed Postgres · encrypted
  • ·9.3M provider graph
  • ·Weekly refresh, nightly DR
/03 · Model
Ranking + next-action loop
  • ·Gradient-boosted ranking
  • ·Feedback-loop fine-tuning
  • ·Versioned, rollback-safe
/04 · Auth
SSO + SCIM
  • ·SAML / OIDC for manufacturers
  • ·Least-privilege by default
  • ·Audit log, queryable
/05 · Integration
API-first architecture
  • ·REST + webhook surfaces
  • ·Warehouse export connectors
  • ·OpenAPI 3.1 documented
/06 · Hosting
US-region cloud · BAA
  • ·Enterprise cloud under BAA
  • ·Region-pinned compute
  • ·Backup retention 35 days

API docs — published at api.solarahealth.com/docs for authenticated manufacturer IT teams.

Security & Compliance

HIPAA, SOC 2, and the boring parts done right.

Solara is built for the regulatory expectations of specialty healthcare manufacturers. The controls are boring because healthcare security should be boring.

/01 · HIPAA
BAA-ready

Business Associate Agreement executed per engagement. PHI handling codified in platform primitives.

/02 · SOC 2
Type II audit track

Control framework mapped to SOC 2 Type II. Third-party audit on a documented cadence.

/03 · Encryption
TLS 1.3 · AES-256

At rest and in transit. Hardware-backed KMS. Documented key rotation.

/04 · Access
Least privilege + SSO

SAML / OIDC enforced. Just-in-time production access, audit-logged end to end.

/05 · PHI Scope
Minimum necessary

Data minimization by design. PHI isolated from analytics surfaces where not required.

/06 · Residency
US regions only

Region-pinned compute and storage. No cross-border data movement without documented exception.

/07 · Incident
60-minute SLA

Documented incident response with a 60-minute first-notification SLA for security events affecting partner data.

/08 · DR
Nightly backup + PITR

Point-in-time recovery for Postgres. DR runbook rehearsed quarterly.

Next Step

Load your target specialty into Solara, live on the call.

Forty-five minutes with a Solara operator. Bring your clinical indications and top geographies — we'll show you the physician universe and day-thirty coverage you'd have.