Resources, Prompts & Datasets
Curated prompt libraries, slide decks from our clinical bootcamp, PhysioNet dataset links, and ready-to-run Google Colab research notebooks for doctors.
Clinical & Vibe Coding Prompt Library
Engineered for instant deployment on Bolt.new, Lovable.dev, Cursor, and Claude 3.7.
12-Lead ECG Waveform Viewer & Arrhythmia Triage
High-contrast clinical ECG canvas with interactive grid scaling (25mm/s, 10mm/mV), automated PR/QRS/QT interval caliper measurements, and simulated arrhythmia flags.
ICU Sepsis Early Warning & SOFA Score Dashboard
Multi-bed ICU telemetry monitor calculating dynamic SOFA, qSOFA, and lactate trend trajectories with hour-by-hour alert heatmaps.
PubMed & STROBE Clinical Paper Drafter Agent
Rigorous scientific research drafting prompt enforcing EQUATOR Network (STROBE/CONSORT) guidelines, statistical reporting rigor, and Vancouver citation formatting.
Radiology DICOM Chest X-Ray AI Multi-Label Classifier
Orthogonal radiological viewer with invert/window-level presets, Grad-CAM attention heatmap overlays, and CheXNet multi-pathology probability bars.
The Zero-to-One Healthcare App Scaffolder
The master prompt from Slide 21 of the MedXchange Bootcamp that structures clinical ideas into buildable, production-ready specifications.
FHIR R4 Resource JSON Schema Generator
Convert clinical notes, lab results, and medication orders into HL7 FHIR Release 4 compliant JSON bundles with SNOMED CT and LOINC codings.
Vibe Coding For Doctors
27-slide curriculum by Faraz Zeeshan (MedXchange Cofounder) & Minaam Qureshi (SDE) · Innovator Circle Bootcamp.
Vibe Coding For Doctors
Innovator Circle Bootcamp | MPM-CX
"Turn clinical intuition into software without getting lost in the syntax weeds."
Doctors hold the deepest domain knowledge in healthcare. Vibe coding provides the leverage to build real clinical software without 4 years of computer science syntax drills.
Clinical AI Google Colab Notebooks
Ready-to-run Jupyter notebooks with pre-configured GPU runtimes, PyTorch, and clinical datasets.
12-Lead ECG Arrhythmia Classification with 1D-ResNet
End-to-end PyTorch pipeline to preprocess raw 500Hz 12-lead ECG signals from the PTB-XL dataset, apply bandpass filtering (0.5-45 Hz), train a 1D Residual Convolutional Network, and compute Macro AUROC & F1-scores for 5 diagnostic superclasses (NORM, MI, STTC, CD, HYP).
Real-Time Sepsis Prediction on MIMIC-IV with XGBoost & SHAP
Extract time-series vitals and lab panels from MIMIC-IV ICU stays using Google BigQuery, perform sliding-window feature engineering, train an XGBoost classifier with focal loss for class imbalance, and generate patient-level SHAP force plots explaining diagnostic risk factors.
Chest X-Ray Pathology Classifier with DenseNet-121 & Grad-CAM
Load NIH ChestX-ray14 / MIMIC-CXR images, fine-tune a pre-trained Torchvision DenseNet-121 with binary cross-entropy loss across 14 radiological findings, evaluate with ROC curves, and generate visual Grad-CAM heatmap overlays highlighting lung opacities.
Clinical PubMed QA with Retrieval-Augmented Generation (RAG)
Build an AI copilot that queries the PubMed Entrez API in real-time, chunks full-text abstracts, generates vector embeddings using BioLinkBERT, indexes into FAISS, and uses LLM generation to answer complex clinical differential diagnosis queries with exact PMID citations.
Survival Analysis & Cox Proportional Hazards for Oncology Cohorts
Perform Kaplan-Meier survival curves, log-rank significance testing, multivariable Cox Proportional Hazards regression modeling, and concordance index evaluation for clinical oncology patient cohorts using the Python Lifelines library.
Medical Named Entity Recognition (NER) with BioBERT
Tokenize unstructured physician progress notes, fine-tune BioBERT for token classification, and extract Diseases, Chemicals, Dosage, and Anatomical Sites with BIO-tagging and clinical F1-score validation.
PhysioNet & Open Medical Datasets
Standard benchmark clinical datasets for ECG, ICU telemetry, Chest X-Rays, and Dermatology.
MIMIC-IV: Medical Information Mart for Intensive Care
The gold-standard global intensive care research database collected from Beth Israel Deaconess Medical Center (2008–2019). Contains comprehensive longitudinal de-identified data including laboratory events, vital signs, medication administrations, ICD-9/ICD-10 diagnoses, procedure orders, and out-of-hospital mortality tracking.
- In-hospital mortality prediction modeling
- Sepsis-3 early warning and vasopressor titration
- Acute Kidney Injury (KDIGO criteria) progression
- Mechanical ventilation weaning predictors
XGBoost, LightGBM, LSTM, Temporal Transformers
MIMIC-CXR-JPG: Chest Radiographs with Radiology Reports
A massive public dataset of de-identified chest radiographs in standard JPEG format associated with free-text radiology reports and 14 structured pathology labels extracted via rule-based CheXpert NLP labelers.
- Automated multi-label thoracic pathology detection
- Radiology report text generation from 2D images (Vision-Language)
- Uncertainty-aware deep learning classification
- Clinical localization and anatomical bounding boxes
DenseNet-121, Vision Transformers (ViT), BioMedCLIP
PTB-XL: A Large Publicly Available 12-Lead Electrocardiography Database
A clinical 12-lead ECG database sampled at 500Hz with comprehensive SCP-ECG diagnostic statements, form statements, rhythm statements, and validated train/test splits. Completely open access with zero credentialing delays.
- Multi-label arrhythmia and conduction disorder classification
- Myocardial infarction (STEMI / NSTEMI) localization
- ST-T change detection and QT prolongation monitoring
- Waveform denoising and generative synthetic ECG generation
1D-ResNet, XResNet1d, Wavelet CNN, Bi-LSTM
eICU Collaborative Research Database
A multi-center critical care database collected through the Philips eICU telehealth program covering 208 community and academic hospitals across the United States. Ideal for testing multi-center generalizability and external dataset validation.
- Multi-hospital algorithmic generalizability testing
- ICU Length of Stay (LOS) prediction
- Telemetry alarm fatigue mitigation
- Resource allocation and mechanical ventilator demand forecasting
Ensemble Gradient Boosting, TabNet, Random Forests
MIT-BIH Arrhythmia Database
The historical foundation benchmark for automated ECG arrhythmia detection, annotated beat-by-beat by two independent cardiologists with over 110,000 labeled heartbeats (Premature Ventricular Contractions, Atrial Fibrillation, Left/Right Bundle Branch Blocks).
- Real-time QRS detection algorithms (Pan-Tompkins benchmarks)
- Single-lead wearable Holter beat classification
- Ectopic beat detection on smartwatch sensors
1D Convolutional Neural Networks, Pan-Tompkins Filter
CHB-MIT Scalp EEG Database
Continuous EEG recordings collected at Boston Children's Hospital from pediatric subjects with intractable seizures, sampled at 256 samples/second with exact millisecond seizure onset and offset annotations.
- Automated epileptic seizure onset detection
- Pre-ictal seizure forecasting algorithms
- EEG channel spatial graph neural networks (GNN)
Spatial-Temporal Graph Convolutional Networks (ST-GCN), EEGNet
ISIC Archive: International Skin Imaging Collaboration
The premier international open-access repository of clinical and dermoscopic skin lesion images with histology-confirmed ground truth diagnoses (Melanoma, Melanocytic Nevus, Basal Cell Carcinoma, Actinic Keratosis, Benign Keratosis).
- Malignant melanoma vs. benign nevus classification
- Lesion border segmentation and 7-point checklist feature extraction
- Smartphone camera dermatological triage apps
EfficientNet-B4, ConvNeXt, Swin Transformer
The Big Prompt Library & GitHub Collections
Community-tested prompt engineering collections, agent system prompts, and UI templates.
The Big Prompt Library
A comprehensive, battle-tested collection of system prompts, multi-persona engineering templates, and coding workflows spanning LLM agents, software scaffolding, and complex technical reasoning.
GitHub Prompt-Library Community Topic
The global curated topic on GitHub aggregating community-driven prompt engineering repositories, LLM jailbreak evaluations, chain-of-thought frameworks, and specialized domain prompts.
MotionSites AI UI Prompt Library
Premium UI/UX prompt library specifically engineered for rapid app generation in Bolt.new, Lovable.dev, Cursor, and v0. Features dark-themed landing pages, interactive micro-interactions, and full-stack components.
MedXchange 10-Week Fellowship
Turn Clinical Datasets into Published AI Research
Learn Python for healthcare, build multimodal foundation models on PhysioNet datasets, and co-author indexed clinical papers with medical peers.