CareFuse

CareFuse

Transparent AI for Hip & Knee Replacement Outcome Predictions. Per-member clinical predictions with SHAP explanations, built for InterQual/MCG workflows and HL7 FHIR PAS.

Visit CareFuse Website

The Perfect Storm: Four Converging Healthcare Forces

The U.S. healthcare system is experiencing unprecedented regulatory convergence that's transforming how knee replacement decisions are made, measured, and governed.

The Accountability Challenge

24%

APU Reduction Penalty for non-compliance with PROMs collection

The Subjectivity Problem

High

Physician Variation - Significant disagreement on TKA indications

The Transparency Requirement

6

States with AI Laws requiring physician oversight of AI decisions

The Evidence Gap Problem

30%

TKAs Avoidable with structured conservative care

CareFuse Solution

Our platform addresses these challenges with transparent, explainable AI that integrates seamlessly with existing healthcare workflows while ensuring regulatory compliance.

FHIR PAS Ready

Built for Da Vinci PAS workflows and ePA compliance by 2027

SHAP Explainable

Transparent AI with per-member clinical explanations

InterQual/MCG Compatible

Seamless integration with existing UM workflows

FDA Compliant

Built under FDA non-device CDS exemption for healthcare compliance

Problem → Approach → My Contribution → Impact

Co-founder & ML Lead at CareFuse

Problem: Payers and providers needed transparent, evidence-based predictions for knee replacement outcomes — with explainability and fit for existing workflows (InterQual/MCG, ePA, audits).

Approach: Build a production ML platform from creating ML-ready datasets through training, validation, deployment, and iteration driven by real healthcare user feedback.

My contribution: Owned the full ML stack: dataset design and pipelines; calibrated models (e.g. logistic regression with propensity weighting) and SHAP explanations; FastAPI + Docker deployment; FHIR R4 and Da Vinci PAS integration; internal tooling and company website; multi-seed holdout validation and FDA non-device CDS compliance.

Impact: AUC ≈ 0.93 with calibration; per-member SHAP explanations for insurer audits; enabled pilot with a major health insurer; platform ready for InterQual/MCG and ePA workflows.

ML & data

  • Developed explainable AI models for TKA outcome prediction
  • Implemented SHAP explanations for clinical transparency
  • Achieved ~0.93 AUC with robust calibration techniques
  • Organized a large OAI dataset for internal model development and analysis
  • Applied propensity weighting for causal robustness

Platform & Deployment

  • Built FastAPI + Docker deployment infrastructure
  • Developed dockerized RESTful APIs for model serving and internal tooling
  • Designed and implemented the company's website and public-facing UI
  • Designed FHIR R4 and Da Vinci PAS integration
  • Ensured FDA non-device CDS exemption compliance
  • Implemented multi-seed holdout validation

Outcomes

Explainable predictions and workflow integration that help providers and payers make evidence-based decisions. We ship, iterate from real feedback, and aim for impact.

~0.93
AUC, calibrated
Pilot
Enabled with major health insurer
SHAP
Per-member explanations for audits