I oversee the domain name search engine and domain parking.

Developed tool to estimate and improve precision of Facebook-related machine learned model. Solved security issue by associating logs with outbound traffic, saving $10k per month. Wrote API connecting Facebook Oauth, public and machine learning data to UI using Hadoop, HBase, MySQL, PHP, Java and Apache.

Wrote ad targeting system using machine learning, Hadoop, SQL, Python, WSGI and Django. My system uses a unique method for targeting that bootstraps company data to find in-segment users. I came up with the concept, created the first pilot code, and supported our first customer with my code.

I developed the People You May Know feature using Hadoop and MySQL. I increased the accept/decline ratio by over 100 percent due to a feedback loop and a linear function. I built several Hadoop clusters from scratch. I used SVM, a machine learning algorithm, to aggregate personal data based on a training set. I wrote the classifier in Perl and Python for this project.

Developed machine learning (SVM) flow for personal profile aggregation. Fed normalized data from Hadoop into both learning and classification steps. Wrote linear classifier in Python and Perl.Designed and wrote back-end People You May Know feature using Hadoop and a linear equation. Captured click through rates and fed data into a feedback loop that raised accept/deny ratio by over 100 percent.Reduced time required for Lucene indexing from 6 months to 30 hours using Hadoop streaming and Python.

Founder of Lunch 2.0. We put techies and companies together for lunch for several years.

Led web spam filtering effort using Hadoop and machine learning to classify and delete pages. Integrated Javascript with crawler. From Hack Day experiment to production, I used a Javascript interpreter, my knowledge of the DOM and the vertical search platform to interpret Javascript in a crawler. This enabled several groups to classify spam web sites and follow deep-web links. Oversaw 1 intern in adult classification. This project employed machine learning techniques (SVM, Bayes) to detect adult text content. Improved throughput of advertising (publisher.yahoo.com) and image (image.yahoo.com) verticals using the crawler to download content and detect 404 errors. Packaged deployment scripts to cut the install time of the vertical crawling platform from 3 days to 20 minutes. I provided needed stability to the platform through health checks.

I investigated crawlers that would be effective for archiving.

I designed 980 nm InAlGaAs lasers for fiber optic communication.