ZX-1 · AI7993 · Spring 2026

AI For Smart
Sensing

ECG heartbeat classification using Retrieval-Augmented Generation and Large Language Models applied to real-world IoT sensor data.

About

Builds a pipeline that converts raw ECG signals into LLM-friendly text, retrieves relevant medical knowledge using hybrid BM25 + embedding search, reranks with a cross-encoder, and classifies each heartbeat as Normal or PVC with step-by-step reasoning.

Team

Matthew McNair

KSU · AI Capstone

Course Info

Course

AI7993 — AI Capstone

Semester

Spring 2026

Project ID

ZX-1

Dataset

MIT-BIH Arrhythmia

Instructor

Dr. Zongxing Xie

🫀 Live Demo

Try the Pipeline

Select a pre-loaded ECG beat, run the simulated RAG pipeline, and watch the system retrieve context and classify the heartbeat — step by step.

Select a Beat

Extracted Features

Mean Amplitude—

Variance—

FFT Mean Magnitude—

True Label—

Select a beat to begin...

ECG Waveform

Retrieved Knowledge Chunks

Run pipeline to see retrieved chunks...

LLM Reasoning

Reasoning will appear here after running the pipeline...

Classification Result

Prediction

—

Confidence

—

Result

—

AI For SmartSensing

AI For Smart
Sensing