EirAI

🌟 Inspiration

SOAP notes are a standardized method of documentation used by healthcare providers to record patient interactions. The acronym stands for:
- S – Subjective: What the patient tells the provider about their symptoms, medical history, and concerns.
- O – Objective: Measurable or observable data such as vital signs, lab results, and physical exam findings.
- A – Assessment: The provider’s medical diagnosis or clinical impression based on the subjective and objective information.
- P – Plan: The treatment strategy, follow-up instructions, prescriptions, tests ordered, or referrals to specialists.
ICD codes (International Classification of Diseases) are critical for billing treatments, but remembering the right code out of over 73,000 is overwhelming.
Nurses and medical coders spend countless hours searching for correct codes just to ensure insurance providers accept a treatment—time that could be spent on patient care.
Doctors face their own bottleneck: writing detailed SOAP notes for each visit, adding to their administrative load.
We spoke directly with working healthcare professionals to validate the need for smarter, AI-powered ICD prediction tools.

EirAI streamlines clinical workflows by reducing documentation and billing friction for nurses, doctors, and coders.
It uses speech-to-text technology to transcribe live patient-doctor conversations into structured SOAP notes.
The notes are then processed to extract the most relevant ICD codes for that patient’s condition.
These codes are cross-referenced with a large dataset of historical claims to identify the ones most likely to be approved by the provider.
The app also promotes transparency in the healthcare system by giving providers data-driven insight into the approval process—helping them avoid denials and navigate the often unpredictable world of medical billing.

Indexed over 73,000 ICD codes for fast and accurate code retrieval.
Trained models on a 1.5 million-row dataset simulating real-world ICD code approval outcomes.
Developed a high-speed vector search system optimized for complex medical queries.
Built a large-scale generative AI pipeline to analyze SOAP notes and recommend ICD codes.
Constructed multiple auxiliary validation datasets to evaluate and fine-tune AI predictions.
Designed a sleek, user-friendly frontend that mimics real-world medical tools.

Realized that even a 1.5 million row dataset only scratched the surface of the complexity in provider-specific ICD approvals.
Capturing nuance across different medical workflows and provider styles was more difficult than anticipated.
Balancing speed, accuracy, and explainability in the AI pipeline presented significant technical hurdles.

Built a fully functional end-to-end prototype that captures speech, generates SOAP notes, and outputs likely-to-be-approved ICD codes.
Successfully integrated multiple AI models, large datasets, and scalable infrastructure.
Applied strong software engineering principles throughout development, even under tight timelines.

Gained deep insight into how real medical professionals manage documentation and billing.
Understood the importance of granular provider-level data in boosting ICD code approval accuracy.
Learned the ethical and technical responsibility of working with sensitive, high-stakes medical data.

Begin partnering with medical professionals across Kansas to gather more targeted datasets.
Scale beyond regional boundaries to make insurance claims more transparent and intelligent nationwide.
Explore EHR integrations to embed EirAI directly into the clinical tools doctors already use.