I've loved computers since I was 5 years old.

Most recently I was the Principal Software Engineer for Microsoft’s Healthcare AI team. My previous company, GitLinks, that I ran as CTO/co-founder was acquired by a multi-billion dollar enterprise software firm. As a 3-time CTO and 2-time founder, I have hired and led engineering teams in 8 different countries and I’ve built and shipped several products from idea to launch and operation.

I’ve founded 2 startups & exited both.

I've had a lot of fun working on Machine Learning/Natural Language Processing (NLP), Architecture/DevOps and UX. Additionally, I'm the inventor on two technology patents.

As CTO and founder of GitLinks I led our acquisition (sold product to Infor). GitLinks was a B2B SaaS product that automated application security and legal compliance for open source software. I built our product, hired and led our engineering team, and pitched technical aspects to investors.

I have a masters degree in Computer Science from Cornell University and a bachelors degree in Electrical Engineering from the University of Houston.

Education

RETINA-AI Health

CTO, RETINA-AI

Automated Diabetic Retinopathy Screening Device. It's "Software as a Medical Device (SaMD)". Our flagship product, the Galaxy™, uses AI to detect diabetic retinopathy from images of the retina. Diabetes can cause blindness, but it can be prevented with early detection and treatment.
 Pending FDA approval.

retina-ai-logo
1 / 2
Pre clinical trial
2 / 2
Setting up the clinical trial

Consulting Services

Consulting

R&D, Software, Technology, Company Strategy, Deck Review, etc.

consulting

Machine Learning Projects

ClinicalBERT: Using a Deep Learning Transformer Model to Predict Readmission

I use ClinicalBERT (BERT pretrained on clinical notes) to predict readmission using only notes from a patient's first few days (not discharge notes). This prediction allows Doctors to actively intervene & prevent readmission while a patient is still in the hospital.

  • PyTorch, Predictive modeling, NVIDIA GPU, AWS SageMaker, Self-Attention, Encoder, Transformer, Transfer Learning, Fine-tuning

Predicting Hospital Readmission using NLP

I built a machine learning model to predict 30-day unplanned hospital re-admission using over 2 million rows of clinical notes (including discharge notes). My approach (AUC=0.83) outperformed the baseline results by 13 points.

  • Healthcare, classification, predictive analytics, hyperparameter tuning, clinical free-text notes
  • AUC, prevalence, recall, precision, ROC, balancing data
  • Scikit-learn, bag-of-words, logistic regression, evaluating performance, learning curve

NLP & Stand-Up Comedy

I walk through text pre-processing and machine learning techniques to answer the question “What makes Ali Wong’s comedy routine stand out?” I do exploratory data analysis on 12 comedian’s routines.

  • Sentiment Analysis — TextBlob
  • Topic Modeling — NLTK, gensim, Latent Dirichlet Allocation (LDA)
  • Exploratory Data Analysis — matplotlib, wordcloud
  • Text Pre-Processing — scikit-learn, pandas, requests, document-term matrix / bag of words

My Startup Work

GitLinks website (acquired)

I wrote this site using Polymer (Google's javascript library for building web applications using web components) and used Firebase's NoSQL Database for storage.

GitLinks web app (acquired by Infor)

I founded GitLinks. GitLinks is an application security, vulnerability management portal that continuously monitors. It supports Java, Javascript (npm) & Python. It integrates with all Continuous Integration (CI) tools like Jenkins, TeamCity, CircleCI, CodeShip etc. Screenshots.

 

Other Projects

Apple watch + iPhone motion authentication

This project uses Swift, Python and a REST API on Heroku. The app verifies a user's identity via motion. The X, Y, and Z g-force measurements from the accelerometer of the iPhone and Apple watch are compared for similarity using a combination of Cosine Similarity and Dynamic Time Warping. Watch video for a match. Watch video for a non-match.

Scoring Engine

In this concept I created a Trust Score (using data science), as a standard in evaluating risk in open source software. The score represented the support level behind an open source project. The 4 categories that made up the score were community support, maturity, popularity and corporate backing.

 

JP Morgan smart contracts on blockchain

I worked with JP Morgan & built a platform for the creation of blockchain-based contracts for trading futures. A future is an agreement between two parties to buy or sell assets at a fixed price to be delivered/paid for later. These smart contracts lived on the Ethereum blockchain & were written in Solidity.

Ethereum

Medical Search Engine

I created a novel medical search engine that could search PubMed based on a combination of CPT code and ICD-10 code mapping (or ICD-9 code) and display the most relevant articles. Read the report or checkout GitHub.

EasyWeave marketplace (Sold)

I founded EasyWeave. EasyWeave is a marketplace platform that connects manufacturers in Asia to buyers across the globe. I created a centralized marketplace with honest reviews. I designed the system architecture from ground up.

Screenshot EasyWeave

Featured in the Press