Skip to content

Software developer builds on artificial intelligence (05/04)

By IAN ROSS For Doug McCormack, writing a data mining software program is one of the more creative technological processes he can think of.
By IAN ROSS

For Doug McCormack, writing a data mining software program is one of the more creative technological processes he can think of.

Doug McCormack
"I feel like I'm writing a song or painting a picture," says the founder and president of CorMac Technologies Inc., a Thunder Bay company specializing in developing custom software in a leading edge field better known as artificial intelligence.

"I just feel creative when I'm working on it. Before you can build something you have to dream it up."

Designing custom software and Web site development pays the bills for the two-person company, which includes his senior programmer Clayton Essex, but their main focus is data mining or expert systems, a process which involves learning to model data and make predictions.

"A word that often fits in describing what I do is data modelling," McCormack says. "Once you can model data, you can get into the optimization of it, and do what-if" scenarios in industrial applications, with stock markets, sports betting, direct mail advertising, employee hiring, property
evaluation and weather prediction.

By optimizing key words such as 'neural networks' and 'data analysis' on his Internet home page (terms most likely to be used by scientists), McCormack relies on Web search engines to bring customers to his virtual doorstep.

It has led to a diverse global clientele that includes the Ontario government, Indiana Medicaid, the U.S. Navy, Swiss pharmaceutical giant Novartis Pharma AG and Alston Power in the U.K.

"We could probably sell more stuff if we did more active marketing," admits McCormack, a 54-year-old, self-taught software programmer who spent a number of years studying techniques to predict the stock market and horse races, intrigued by "the mathematics of it all."

With a civil engineering background, he worked for 15 years in the forest products industry in Ontario and Alberta, where he wrote a log-sawing program on how to optimize cutting lumber from logs.

"There's not a lot of difference between building a bridge and a software program. You have to figure out what you need and get all the components with the right strengths, in the right places or the whole thing doesn't work."

He returned to his hometown of Thunder Bay and opened a computer store in 1987, but found retailing not to his liking.

After taking a Confederation College night course, he learned about artificial neural networks - "the cutting edge of data analysis" - a computer program that learns to recognize patterns by examples rather than just spitting back pre-programmed answers.

He sold the store in 1994 to start CorMac Technologies and wrote the first version of his NeuNet (Neural Network) data mining prediction program a year later.

He started selling it on CompuServe, the bulletin-board precursor to the Internet and the product immediately took off, selling packages to individuals predicting the outcomes of greyhound dog racing, doing cancer research in South America and for an insurance firm using it as a tool for analyzing
insurance claims.

Some U.S. companies are using these forms of AI to expose credit card fraud and the medical and scientific research industry is starting to catch on.
The U.S. Center for Disease Control in Atlanta uses McCormack's VisiRex (Visual Rule Extraction) program to study what combinations of factors contribute to adverse reaction to vaccinations.

His latest project is in the biotechnology field, "dabbling" with Thunder Bay's Genesis Geonomics Inc., analyzing DNA sequences they are collecting from cancer patients.

Several universities around the world are linking to his NeuNet collection for their students to download and use for research.

Along the way the company has developed some product offshoots such as their text-to-speech software.

Together with his brother Rob McCormack, a Confederation College professor, they developed a text-to-speech reading machine program for their father who was losing his sight and could not read a newspaper.

It led to a spin-off company (Readplease.com) selling screen reading software program that reads Web pages and e-mails.

The U.S. Navy eventually downloaded a free version of Readplease and contacted McCormack to custom-build them a version for their war game scenarios.

CorMac Technologies developed a multi-language custom text-to-speech application now used by as a battlefield voice-simulator training tool with the USS Enterprise carrier battle group.

McCormack says he is now focussing his energies on updating his NeuNet and VisiRex programs by incorporating some classical statistical tools such as multi-variable regression and re-packaging them into a more advanced, high-end product.