** Feb 2, 2026 — Project Started**
We kicked off Discharge Risk AI to address a persistent hospital operations problem: discharge delays that occur after patients are medically ready to leave. The goal is to predict discharge risk earlier and explain why a discharge is likely to fail before it becomes a bottleneck.
** Feb 3, 2026 — Discharge Risk Scoring Implemented**
Implemented the first version of the discharge delay risk scoring logic using structured, non-clinical operational inputs such as insurance type, placement constraints, and timing factors.
# Simplified example of risk scoring inputs
features = {
"payor_type": payor_type,
"snf_availability": snf_beds_available,
"distance_to_placement": miles,
"authorization_required": prior_auth,
"day_of_week": discharge_day,
}
risk_score = xgb_model.predict_proba(features)
Early testing confirmed these signals can meaningfully predict discharge risk without using PHI.
** Feb 4, 2026 — Backend API Deployed on Google Cloud**
Built a FastAPI backend and deployed it on Google Cloud Run, establishing a production-grade API layer and clean separation between inference and presentation.
@app.post("/assess-risk")
def assess_risk(payload: DischargeScenario):
score, factors = calculate_risk(payload)
explanation = explain_with_gemini(score, factors)
return {
"risk_score": score,
"factors": factors,
"explanation": explanation,
}
** Feb 5, 2026 — Gemini 3 Explainable Reasoning Added**
Integrated Google DeepMind Gemini 3 as the explainable reasoning layer. Gemini interprets structured model outputs to produce clear explanations and operational recommendations.
prompt = f"""
Given a discharge risk score of {score}
and these contributing factors: {top_factors},
explain why this discharge is at risk and suggest
early operational actions. Do not provide medical advice.
"""
Gemini is used for reasoning and explanation, not prediction.
** Feb 6, 2026 — Public Web App Launched**
Launched the first public version of the web app using React and Firebase Hosting. The demo is intentionally login-free so judges can test it immediately.
<button onClick={handleAssessRisk}>
Get Risk Assessment
</button>
(Attach dashboard screenshot here)
** Feb 7, 2026 — Explainability & Dashboard UX Improvements**
Refined the dashboard to emphasize the discharge risk score, ranked contributing factors, and Gemini-generated explanations. Improved layout and hierarchy to resemble an enterprise healthcare operations interface.
<RiskGauge score={riskScore} />
<FactorList factors={topFactors.slice(0, 3)} />
<GeminiExplanation text={explanation} />
(Attach explainability / risk visualization screenshot here)
** Feb 8, 2026 — Visual Identity & Architecture Added**
Introduced a minimal logo mark and subtle dashboard hero visuals to reinforce product credibility. Finalized a one-page architecture diagram showing Firebase Hosting, Cloud Run, XGBoost risk scoring, and Gemini 3 explainable reasoning.
Browser → Firebase Hosting → Cloud Run API
→ XGBoost Risk Model
→ Gemini 3 Reasoning Layer
→ Explainable Results UI
(Attach logo + architecture image here)
** Feb 9, 2026 — Demo Video & Final Submission Ready**
Recorded a short demo video showing live risk assessment, contributing factors, and Gemini explanations. Completed final QA, verified public access, and prepared Discharge Risk AI for submission to the Gemini 3 Hackathon.
Current Status
- Fully deployed on Google Cloud
- Public, login-free demo
- Explainable AI powered by Gemini 3
- Hackathon submission complete
Log in or sign up for Devpost to join the conversation.