Senior Machine Learning Engineer at Bio Conscious Technologies
Originally from Moscow, Russia, Stan has led the Bio Conscious machine learning department since 2018.
Stan is a science and math enthusiast. He started his post-secondary academic career at the Moscow Engineering Physics Institute with a BSc in solid state physics and went on to continue his academic career at Pennsylvania State University, where he got a MSc and PhD in electrical engineering. While working as an assistant professor teaching classes in electrical engineering at Tulane University, Stan obtained yet another MSc, this time in computer science with an emphasis in artificial intelligence.
Stan is a man of many talents. During his education he pursued a career in both professional chess and poker, where he won over $1,000,000 in online cash games. As such, Bio Conscious has a strict standing policy that no poker is to be played at company events, especially if Stan is in attendance. Chess is allowed, but there are few challengers.
Beyond his skills in math and science, Stan is also an accomplished singer, and regularly takes part in choir performances. He also has a fondness for the great outdoors. “I love the mountains and hiking”, he replied when asked about living in Vancouver, “also, it’s not as cold as Toronto or Montreal.”
At Bio Conscious, Machine Learning is the core of what we do. We’re continually adjusting and developing algorithms in the pursuit of finding the most accurate and user friendly method of predicting blood glucose levels in patients with diabetes.
“The hardest part about my job is figuring out how to deal with bad data”, Stan says, “Unfortunately, most people don’t tell us when they do things that change their blood sugar. The algorithms think that the user’s body behaves a certain way naturally, when in reality they’ve taken insulin or eaten carbs, causing that change. The data is incomplete, and we need our models to account for this.”
It turns out, predicting blood sugar levels is relatively easy when you have perfect data. It’s when working with unstructured and incomplete data that the rubber really meets the road and we find out how good our models really are.