The Challenge

After much reading about Vitamin D, the idea was hatched to make an app that would help people keep on top of how much they were getting when getting safe sun exposure, and what their likely level was.

The Solution

Make an app that could get satellite information from the phone to figure out where on the planet the subject was and construct the solar forecast, then make it possible to track how much D (s)he was getting for each minute outside and estimate how much it would move the overall level.

Not long after it was released, it came to the attention of the world's leading authority on Vitamin D, Dr. Michael Holick, and a partnership was forged.

The Challenge

Make a tool that people could carry around that would help them zero in on things to do when traveling.

Using web or the native tablet app, people could find things based on their interests and throw them into an itinerary that they could work on before taking a trip.

The Challenge

At the peak of the government's putting contracts and grants out for bids, as many as 5000 new opportunities show up per day. Especially small players don't have time to go through all them to see if there are items they might be interested in.

The Solution

A (Bayesian) predictor that scores each opportunity for each member, making sure that time is spent on the most valuable items first.

The Result

System was originally developed for Contracts and was then adapted to Grants and has been in production ever since.

The Challenge

To provide an end-to-end system for running the employee satisfaction practice.

The Solution

A portal for the management of clients, reports, etc. System for generating books of reports from client data. Still in use.

The Technology

Java EE solution built with Hibernate and Spring.

The Challenge

To rewrite a C codebase in Java, increasing interoperability and performance.

The Solution

Implementation of a Bayes inference engine that classified typing samples by characteristics like flight time, hold time, etc. Logic was introduced that allowed us to figure out who was authenticating before the credential typing was complete so that we were not loading the sample data while the system/user were waiting for a decision.

The Technology

Java

The Challenge

Company had a huge list of people to call and a short amount of time to sign up as many people as possible.

The Solution

We wrote a predictor that each night would train on all the to date results, then predict the likelihood of interest of everyone remaining to be called so that focus was constantly on the people who were most likely to say yes.

The Technology

Java