Alberta, and Edmonton, in particular, has a strong Artificial Intelligence (AI) and Machine Learning (ML) sector, which many of us hope will continue to thrive. In order for us to sustain and grow this section of technology, we need leaders willing to identify areas of potential disruption using AI/ML and to take the risk to build out models.
John Murphy, CEO of Stream.ML is one of those risk takers. John started his first tech company in the ‘80s growing it from two people to more than 100 before successfully exiting. That company, Shana Corp, was purchased by Filenet, which was then purchased by IBM. An entrepreneur at heart, John has experienced both sides of the startup world with some good wins and a few fails, but that hasn’t dampened his enthusiasm for technology. He is constantly exploring new ways to use it.
When it came to Stream.ML, John and his team were looking at how to bring the world of chemistry and digital devices together in a way that provides more immediate access to analytics. Digital devices, whether cameras, spectrometers or sensors, collect data from chemical to visual to molecular. John’s team felt that there should be a way to take that data and apply machine learning so it could be analyzed more efficiently and cost effectively.
When the team at Stream.ML started investigating this, they enlisted the University of Alberta and determined that building a single model could cost upwards of $100,000 and take up to eight months. In terms of building a scalable platform and being able to work with multiple devices, this potential cost was prohibitive and John and his team wondered whether there was a way to make machine learning more accessible.
This led to the development of Stream.ML’s platform, which has been designed to allow easier building using autoML and incorporate Application Programming Interfaces (API’s) for integration to existing software and hardware solutions.
By building a platform and using tools like Google and Microsoft for machine learning Stream ML allows users to upload their data, whether text or visual, and build their own models. For example, it could begin with the user uploading and labeling the data, labeling a picture of a plant with a specific disease, identifying how much protein is in barley and what that means, or correlating the blood pressure of a client with a specific health issue. Then with a bit of drag and dropping, the user can push a button and get an AutoML model. This can be done in as little as an hour, and from there can be executed in a second.
An example from a recent project is based around cannabis. Imagine you are walking through the greenhouse and you see an issue on a plant. As a worker, you could open up an app and take a picture of the plant in question. It could then return to you an answer on what type of issue you are seeing. This is what Stream.ML makes possible. The worker doesn’t have to know anything more than how to use the app to get access to years of data and instant analysis—allowing them to identify an issue and report and/or address it right then.
It took time and experience to get to this level of analysis and capability. John’s team started by taking images and identifying what a person could see with their eyes. As they gained experience, they were able to extend this out to include device information beyond camera images to multispectral images (image data within specific wavelength ranges across the electromagnetic spectrum) and spectrometer data (which measures frequency and energy of light). Combining this data with neural nets allows the identification of nutritional levels or disease.
This approach can be used in agriculture, health care and food technology. As Stream.ML was developing its platform and discussing its approach with students, professors and industry, they had a lot of interest from people trying to build a model. As a result, Stream.ML has opened up its platform for users to create an account and upload their own data to make a model. Whether you are a student who wants to give it a try, or you are a company that can see how this would help your business, you can try it out at no cost. John and his team are opening up a marketplace so that you can potentially share or sell your models.
When I asked John “Why Alberta”, it was a simple answer. He’s from here, and Alberta has treated his company well. Beyond that, John stays here because Alberta has unbelievable tech, universities and colleges that are producing great grads. We also have great foundational tech being built. Our community is coming together as an ecosystem and John envisions world-class tech being built here as we have everything to get it done.
Check out our full interview to learn more:
Interested in what else John is involved in? Check out our blog on Bio-Stream Diagnostics, a company that is working towards a 30-second COVID Test.