ShifaMind is live · concept-grounded ICD-10 coding

AI infrastructure for clinical reasoning.

Roshan AI builds clinical-grade models that explain themselves. The same concepts a doctor reaches for, surfaced as verifiable evidence on every prediction. ShifaMind is the first product on the platform.

Products
Layer 04
ShifaMind · NabzGraph · future products
Reasoning
Layer 03
Concepts · explanations · evidence
Models
Layer 02
Clinical encoders · concept bottlenecks · decoders
Ingestion
Layer 01
Notes · structured records · imaging

Accuracy comparison

Macro-F1 on MIMIC-IV top-50 (higher is better)
MIMIC-IV
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0.164
0.343
0.417
0.435
0.649
0.712
Vanilla CBM
Claude 4.6
GPT-5.4
Gemini 2.5 Pro
GKI-ICD
ShifaMind
ShifaMind
Benchmarks

Ranked #1 on automated medical coding.

Highest Macro-F1 across frontier general-purpose LLMs and the latest published clinical-coding work.

0.712
Macro-F1, highest in class.

Beating Anthropic (0.343), OpenAI (0.417), and Google (0.435) on the same MIMIC-IV top-50 ICD-10 evaluation.

>60%
Improvement vs. the best general-purpose LLM.

Frontier models stretch into healthcare; ShifaMind is built for it. Same evaluation, ~0.28 absolute Macro-F1 lead.

#1
Concept-grounded by construction.

The only architecture in the comparison that produces verifiable clinical-concept evidence alongside every prediction.

Why Roshan AI

Clinical AI that earns the trust to be deployed.

Clinical-grade reasoning

Models trained on real clinical data and structured against the concepts a doctor uses. Not a general-purpose LLM stretched into healthcare.

Built for integration

API-first from day one. Drop predictions, evidence, and concepts into existing workflows (EHR, coding tools, dashboards) without custom integration work.

Evidence on every prediction

Each output carries the activated clinical concepts and the verbatim phrases in the note that produced it. Every prediction is defensible.

Compliance posture

HIPAA-ready deployments, encryption everywhere, audit trails on every inference, and a strict no-training-on-customer-data default.

Roadmap

One platform. A growing line of products.

Every Roshan AI product runs on the same ingestion layer, the same clinical encoders, and the same concept reasoning layer. New products and capabilities slot into the stack without bespoke plumbing. Here is what is live today and what is coming next.

What's next on ShifaMind
Next
Capabilities in active development.
  • Multi-agent reasoningIn dev
  • HCC codingIn dev
  • Risk Adjustment Chart ChaseIn dev
Further out
Later
More products on the same stack.
  • New clinical-reasoning productsPlanned
    Same ingestion, encoders, and reasoning layer
Clinicians · developers · partners

Build with Roshan AI.

Pilot ShifaMind, integrate the API, or talk to us about the next product on the platform. We work directly with the people who ship the workflow.