Greg Stewart

Greg StewardAdjunct Professor
BASc (Dalhousie), MASc (Dalhousie), PhD (UBC)







I have a B.Sc. in Physics and a M.Sc. in Applied Mathematics from Dalhousie University, and a PhD from University of British Columbia in Control Engineering. Fellow of the IEEE and IFAC, received the IFAC Industrial Achievement Award 2017, twice received the IEEE Control Systems Technology Award in 2002 (papermaking) and in 2012 (automotive), the IEEE Transactions on Control Systems Technology Outstanding Paper Award. 46 patents, more than 57 technical publications, and designs residing on over 300 industrial installations.

I enjoy mountain biking, cooking, and am a certified judge for competitive barbecue.

Research Interests

My passion is in developing and deploying advanced machine learning (or AI?), control, and automation technologies into commercial production to address meaningful problems.

I am fascinated by the challenges and opportunities posed by agricultural problems. Compared with domains such as manufacturing, agriculture has similarities (eg desire for uniformity, predictability, and profitable throughput) and some important differences (eg biological uncertainty). It is a field (pun intended) that is starting its journey to benefiting immensely from measurement, analysis, and automation technologies that have revolutionized other domains. While working as VP Data Science with Ecoation I led the development and productization of:

  • Production-deployed models include several industry-firsts for agriculture – tomato count, pepper count, pest and disease projection (link), detection of microclimates, and crop work. Modeling technologies include neural networks (vision), epidemiological modeling, anomaly detection (link).
  • First commercial modeling of multivariable interactions among these variables and providing guidance to users on recommended actions such as pest treatments (link, link).

These machine learning products resulted in amassing a valuable data set unique in agriculture and 7 patents filed.

During the pandemic, together with Klaske van Heusden and Guy Dumont we have been working to develop and communicate the concept of feedback in the development of models and public policy setting. Feedback has the potential to reduced the large uncertainty in projection accuracy by an order of magnitude (links to article, interview).

Along the way I have been lucky enough to work with a wide range of industries – papermaking, automotive powertrains, semiconductor manufacturing (link), carbon fiber airplane brake production, combustion health in industrial burners, large scale data centers, microalgae cultivation, reinforcement learning for maintenance-free industrial control (link, link). These machine learning and control projects have resulted in more than $30M/year in value, designs residing on over 300 production installations, and the standing up of a new business – Honeywell Automotive Software (now Garrett Motion link to recent exciting news about MPC in production).