Medical Intelligence · Open Source

ICD-10 Coder
for Insurance Intelligence

A fine-tuned Qwen2.5-7B model that decodes WHO medical classifications into insurance coverage estimates. The first agent in a multi-scheme, MoE pipeline built to serve the Indian healthcare ecosystem and beyond.

8B
Parameters
A5000
Trained on
2h
Training time
Faster via Unsloth
WHO
ICD-10 Standard
Input
Clinical text
ICD-10 Model
Qwen2.5-7B
Web Search
Tariff lookup
LLM Analysis
Claude / Gemini
Output
Coverage report
Model & Backend
Primary ICD-10 Model
Coverage Analysis LLM
Integrations
Web Search
Real-time hospital tariff & pricing lookup
Coverage Estimation
WHO-tier reimbursement brackets
Admission Rules
Min. days, pre-auth, eligibility flags
About AxisMapper

Insurance agents, hospital billing teams, and patients have no reliable way to know what a given diagnosis actually entitles them to. Coverage decisions are opaque, rules fragment across schemes, and the same condition can be coded five different ways — each triggering a different payout.

AxisMapper is being built to decode that opacity. This is Phase 1: a WHO ICD-10 base model that understands international classification standards and applies them to coverage logic.

Future phases will fine-tune separate expert models for Ayushman Bharat, Star Health, ESIC, CGHS, and state-level schemes — each with its own tariff rules — then route queries through a Mixture-of-Experts router that picks the right specialist for each patient's insurer.

GitHub ↗ Model Card ↗
Query Interface
Enter a clinical scenario
Quick scenarios
📋 Analysis Results
Code Type Description Chapter
WHO-tier coverage estimate
🌐 WHO ICD-10 + Web tariff data
    🗺️ AxisMapper Roadmap
    ✅ Phase 1 — Now
    WHO ICD-10 Base
    • Universal code prediction
    • WHO-level coverage logic
    • Admission rule extraction
    • Multi-LLM analysis layer
    • Web tariff search
    🔧 Phase 2 — Next
    Scheme-specific Models
    • Ayushman Bharat / PM-JAY
    • Star Health rules
    • ESIC / CGHS tariffs
    • State scheme variations
    🔀 Phase 3 — Planned
    MoE Router
    • Insurer identification layer
    • Specialist model routing
    • Confidence-weighted outputs
    • Ensemble code resolution
    🤖 Phase 4 — Vision
    Multi-Agent Pipeline
    • Agent 1: Diagnosis → Code
    • Agent 2: Code → Coverage
    • Agent 3: Rules → Claim amount
    • Agent 4: Web validation