Scarlat Machine Intelligence


Web apps

cda2sqllogo-1 cda2sql – App & API that parses C-CDA documents into SQL tables
cropped-logo300x300.png – extracts over 5.7M medical concepts and their Unified Medical Language System codes from free text, over the web

Text2Codes – extracts medical terms and physician fees from free text for billing purposes

Screen Shot 2018-09-05 at 4.24.52 PM.png – identifies a dog breed in an image with 97% accuracy
Screen Shot 2020-07-19 at 1.42.17 PM Symptoms, signs and their differential diagnosis


Screen Shot 2018-08-18 at 11.27.28 PM.png Hand written digit recognizer  Top 2% Kaggle competition July 2018
Screen Shot 2018-08-18 at 11.32.14 PM.png

House prices prediction Top 3% Kaggle competition July 2018

Projects / Public Kernels

Screen Shot 2019-09-16 at 4.34.10 PM.png Future sales forecasting
Screen Shot 2019-07-11 at 11.57.33 AM.png Unsupervised detection of opiates diversions in a hospital
Screen Shot 2019-06-11 at 1.47.14 PM.png Predict antibiotic resistance with bacterial gene sequence -with 99% accuracy
Screen Shot 2018-08-18 at 10.59.07 PM.png Compare 6 unsupervised anomaly detection algorithms identifying fraud in credit card transactions
Screen Shot 2018-11-25 at 3.42.37 PM.png Identify Melanoma in Images with 95% accuracy
tenor.gif Predict Hospital Mortality of ICU Patients (MIMIC3)
Screen Shot 2018-11-04 at 6.38.20 PM.png Predict Hospital Length of Stay LOS  (MIMIC3)
Screen Shot 2018-10-14 at 5.28.52 PM.png Weather temperature prediction time sequence analysis
Screen Shot 2018-08-19 at 12.20.42 AM.png IMDb sentiment analysis – predicting movies’ ratings based on their text reviews with supervised binary classification
Screen Shot 2018-08-19 at 12.18.18 AM.png Document classification – Reuters news
Multi-class, mutually exclusive, classification of free text news
Screen Shot 2020-03-02 at 3.39.19 PM.png Compare 10 unsupervised clustering algorithms on the Iris dataset
Screen Shot 2018-08-19 at 12.14.09 AM.png Compare 14 linear regression models with a NN Cross validation, model evaluation, hyper-parameters optimization
Screen Shot 2018-08-19 at 12.09.26 AM.png MNIST digit identification
Old challenge and some new tricks – 99.75% accuracy identifying hand written digits
Screen Shot 2018-08-19 at 12.22.28 AM.png Feature engineering via pipeline
Model learning curves, control under and overfitting

Machine Learning Primer for Clinicians

Screen Shot 2018-12-20 at 11.20.52 PM.png

A series of articles on ML with no math and no programming,  published on HIStalk blog

The paperback is now available on Amazon

  1. Introduction to ML
  2. Supervised Learning
  3. Unsupervised Learning
  4. How to Properly Feed Data to a ML Model
  5. How Does a Machine Actually Learn
  6. Artificial Neural Networks Exposed
  7. Controlling the Machine Learning Process
  8. Predict Hospital Mortality
  9. Predict Hospital Length of Stay
  10. Anomaly Detection in Antibiograms
  11. Basics of Computer Vision 
  12. Identify Melanoma in Images


Screen Shot 2018-08-19 at 12.25.47 AM.pngScreen Shot 2018-08-19 at 12.27.54 AM.pngScreen Shot 2018-08-19 at 12.28.27 AM.pngScreen Shot 2018-08-19 at 4.54.08 PM.pngScreen Shot 2018-08-19 at 4.59.04 PM.png

Screen Shot 2018-08-19 at 12.26.56 AM.pngScreen Shot 2018-08-19 at 12.29.16 AM.pngScreen Shot 2018-08-19 at 12.25.07 AM.pngScreen Shot 2018-08-20 at 2.21.19 PM.png

Screen Shot 2018-08-19 at 12.24.39 AM.pngScreen Shot 2018-08-20 at 2.22.07 PM.pngScreen Shot 2018-08-20 at 2.20.49 PM.pngScreen Shot 2018-08-19 at 12.23.21 AM.png

Python, Octave, Matlab, SciPy, Scikit learn, Pandas, Matplotlib, NumPy, TextRazor, Jupyter, RapidMiner, NLTK, Keras + TensorFlow

Image: Salvador Dali (1951): Raphaelesque Head Exploding

Updated October 4th, 2020