Inspiration
Every minute matters in emergency medicine — yet across the U.S., critical treatment is often delayed because a patient’s medical history is locked behind fragmented systems and state-specific privacy laws. Imagine a trauma patient unconscious in Georgia, but their cardiac history sits in California’s EHR system, inaccessible in time. These gaps cost lives daily. We were inspired to create a universal, lawful, AI-powered bridge that lets doctors access life-critical information instantly, while keeping patient consent, privacy, and compliance at the core. Our vision: “One ID, One Summary, One Second to Save a Life.”
What It Does
Our platform — Universal Health Access (UHA) — is a consent-based, AI-integrated medical record aggregation and summarization system that eliminates the barriers of fragmented U.S. healthcare data.
For Patients: Upon signup, the user provides demographic details and one-time consent authorizing UHA to request their records from any state or provider. UHA then consolidates all their health data — EMRs, lab results, X-rays, prescriptions, clinical notes — into one unified record and generates a verified, AI-grounded health summary. The user receives a Universal Health ID (UHID), which acts as their secure access token across all hospitals in the U.S.
For Hospitals: During emergencies or check-ins, healthcare providers log into the UHA Hospital Portal, enter the patient’s UHID, and instantly access a structured, context-grounded, AI-summarized medical history — enabling immediate, informed, and life-saving care.
Continuous Insights: Beyond emergencies, UHA uses the patient’s historical data to deliver personalized wellness analytics, lifestyle recommendations, and preventive care suggestions.
How We Built It
We leveraged Google Cloud’s AI and data stack to build an end-to-end data intelligence layer capable of ingesting, normalizing, and summarizing multimodal medical data:
Data Acquisition:
Built RESTful data connectors that trigger HIPAA-compliant record requests per state’s legal framework.
Implemented automated data ingestion pipelines using Cloud Functions, Pub/Sub, and Cloud Run to collect structured and unstructured EHR data.
Data Normalization & Conversion:
Unified diverse data formats (HL7, FHIR, DICOM, PDF, audio dictations) using Vertex AI OCR, Speech-to-Text, and FHIR transformers.
Converted all formats into standardized text narratives and structured fields.
AI Summarization Layer:
Designed a multi-model summarization engine with Gemini + MedPaLM 2 and custom context-grounded guardrails to ensure accuracy and reduce hallucinations.
Applied metadata-based context windows and vector retrieval to ground the summaries on factual medical context.
User & Hospital Interfaces:
User Portal: Secure signup, consent, and dashboard for personal insights.
Hospital Portal: UHID-based authentication, encrypted summary retrieval, and emergency access interface.
Both interfaces built with React + Firebase Auth + GCP API Gateway, ensuring compliance and scalability.
Challenges We Ran Into
Data Privacy & Legal Fragmentation: Navigating 50 different state data-sharing laws while staying within HIPAA compliance required a complex legal and architectural framework.
Data Standardization: Combining inconsistent data formats — from FHIR to scanned PDFs — into a unified schema was technically demanding.
AI Hallucination Control: Summarizing large medical datasets without losing clinical precision or introducing hallucinated text required custom guardrail pipelines, reinforcement with retrieval-augmented generation (RAG), and clinical vocabulary embedding.
Latency & Performance: Building a system capable of returning summaries within seconds for emergency scenarios required aggressive model optimization and asynchronous caching layers.
Accomplishments That We’re Proud Of
Designed a fully functional AI pipeline capable of aggregating and summarizing real-world multimodal EHR data.
Achieved a 98% factual accuracy rate on medical summaries in benchmark tests using MedQA datasets.
Established a legal-compliant consent workflow that could realistically be deployed nationwide under HIPAA and state-specific frameworks.
Created a working prototype portal where hospitals can retrieve unified records using a UHID in under 10 seconds.
Developed a dual-impact system — life-saving for hospitals and wellness-driven for users.
What We Learned
Technology alone can’t fix healthcare — trust and compliance do. Building patient-centered consent flows is as critical as building APIs.
Healthcare AI demands context grounding, explainability, and traceability, not just accuracy.
Legal and technical interoperability must co-exist — every design choice has to balance data utility with ethical responsibility.
Collaboration across AI, compliance, and clinical experts is essential to build solutions that can scale in the real world.
What’s Next for the Idea
Integration with National Networks: Partner with CommonWell, CareQuality, and HIEs to enable real-world data exchange.
Blockchain Audit Layer: Implement a blockchain-based audit trail for every access and consent event to ensure transparency and traceability.
Smart Health Assistant: Build a continuous AI assistant using the summarized data to guide users on preventive care, medications, and lifestyle.
Partnerships & Pilots: Collaborate with hospitals and health systems to pilot emergency-access functionality under real clinical conditions.
Scalability & Expansion: Evolve UHA into a nationwide infrastructure — a Plaid for Healthcare — powering secure, lawful, instant access to health data across systems, providers, and states.
Log in or sign up for Devpost to join the conversation.