Mara Cairo, Product Owner of Advanced Technology at Amii – Interview Series

Mara Cairo is passionate about using AI for good. She has a Bachelor of Science in Electrical Engineering from the University of Alberta and holds her P.Eng. and PMP designations. Before joining Amii, she worked in the hardware development space, where she helped clients take their products to market, with a focus on micro and nano-fabrication.

As Product Owner of Advanced Technology at Amii, Mara leads a technical team that helps industry partners build machine learning capacity within their organization by providing guidance and expertise to develop predictive models. Her team works with clients who are committed to advancing along the AI adoption spectrum by applying machine learning to their most challenging business problems.

Amii (Alberta Machine Intelligence Institute) is one of Canada’s preeminent centers for AI, they partner with companies of all sizes, across industries, to drive innovation strategy and provide practical guidance and advice, corporate training and talent recruitment services.

We sat down for an interview at the annual 2023 Upper Bound conference on AI that is held in Edmonton, AB and hosted by Amii.

What initially attracted you to electrical engineering?

As a kid, I just really liked building things. My mom would bring home a fan when it was hot in summer, and I would want to build it. I remember growing up as a teenager, I had a cell phone, one of those Nokia’s that you could take apart and I would take it apart and put bejewels all over it on the inside and the antenna. But when I opened it up, it was like, “Holy crap, what’s in here? What’s going on?” It was really interesting to me.

I always excelled in math. So, putting all of those together, my parents also pushed me in the engineering direction because I was good at math, I had just a general interest in electronics and wanted to know more about it, that’s kind of what drew me in to begin with.

Also, in engineering, I just really liked the idea of applying math to real-world problems. Yeah, okay, cool, math is great and exciting and fun for me, but with engineering you can apply it to solve hard problems. It seemed kind of the perfect mesh of things that would lead to an interesting career.

Your parents sounded very proactive in supporting your interests.

Yeah. My dad especially. He says he saw it in me from a young age and just always pushed me in that direction. I was at a Women in AI event last night too and we talked about removing some barriers and making it a more approachable field for women. And I didn’t really see that as a barrier because, again, my parents were like, “This is what you should do. It’s not a question of your gender or anything. It’s just this is a skill you have. You should naturally kind of follow it and nurture it.” I never felt like it wasn’t for me, which helped obviously.

Before joining Amii you worked in the hardware development space to focus on micro and nanofabrication. Could you define those terms?

Definitely. So, in electrical engineering, I took the nanoengineering option. It was the specialty around designing and manufacturing on the micro and nanoscale. When we talk about a nanometer, we’re talking about a millimeter divided in a million is a nanometer. A very, very small scale. And that’s cool. These things are so small you can’t even see them with the naked eye. But I could take this specialization to learn how to manufacture on that scale and design things on that scale.

We live in a very connected world. There’s electronics all around us and we need to be able to design electronics for the packaging and space constraints. We’re constantly trying to make things smaller and smaller. You take something bulky, a prototype, and you need to be able to make it reproducible and scalable. Nanofabrication is really about the tools and the techniques that you use to design and manufacture on that kind of level.

This is from manufacturing microchips to taking those two different chips and connecting them electrically to the final packaging. Doing all of that on the microscale requires a different technique than building something on our human scale. The micro and nanofabrication are just around the chemical processes that you use and the electrical processes, the packaging that you need to make sure these are hermetically sealed and protected from their environment.

Outside of microchips, what would be another application or use case?

We worked on a lot of projects like fiber optics. Again, it all eventually must come to some sort of processing unit that’s taking in signals or generating signals. We did work in the telecom industry, optics, cameras, all of that stuff. But the brains of it are generally some sort of microchip in the middle. But there’s also the sensors that are feeding their signals into whatever processing unit you’re using. So diverse manufacturing techniques for building whatever type of sensor or input or output device that we need.

What are some of the challenges behind working on this type of nanoscale?

One piece of dust can ruin your whole day. Things you’re working on are the same size as the dust in the air. So, you fabricate in a clean room. The clean room is really an environment that’s protecting what you’re working on from you as a human, because we are very dirty as humans, we’re constantly kind of spitting out particulates, our clothes are particulating, the makeup that we’re wearing it’s making the air dirty. We need to eliminate as much of that as possible so that the things that we’re building are clear and clean of that sort of contaminant.

Another challenge, there’s great ways to build these clean rooms and there’s a whole kind of study and science behind that, but the other challenge is taking it out of the lab because eventually these things are going to be used in our very dirty world. That’s when the packaging becomes important. We still need to be able to access these devices, but we need to do it in such a way that we’re not contaminating the environment, the packaging. So hermetically sealing things, making sure it’s completely sealed, nothing’s getting in or out. That’s another set of challenges that I saw. We would have something that works great on a lab bench in a controlled setting, but generally most of the things that we’re building are meant to be brought out into our dirty world. That was challenging as well.

Again, from manufacturing all the way to taking it to its final destination, it’s just very special kind of considerations and environmental concerns when you’re dealing with things that small. Also, things don’t always behave as expected on that small of a scale. In our physical world, we expect things to work a certain way, but when you get down to the micro and nanoscale, the physical world becomes a little bit different, and you can’t always anticipate the results. That’s a whole other field of study.

What would be some examples of being different than the regular physical world?

Passing current through a wire. We have our chargers and our phones and we’re passing current through it. When you’re passing current through a wire that’s sized like a strand of hair, there’s obviously heat considerations and things will just start behaving differently because, again, the space and the size constraints.

What is your current role at Amii, and how does your team help industry partners?

My current role at Amii is vastly different from the world of micro and nanotechnology.

I’m Product Owner of the Advanced Technology Team at Amii. I lead a team of mostly machine learning scientists and project managers who are all working with our different industry partners to solve their business problems through the application of machine learning.

We’re very industry-focused, all about bridging the gap between what’s happening in academia, all of the really great breakthroughs with machine learning and AI but applying them to our industry partners biggest needs. We respond to those needs by essentially helping our clients find the skills and the expertise that they need to be able to move the work forward.

We run our internships and residencies program through the advanced technology team. So, I’m hiring a lot. Recruitment is not my background, but it’s something I do a lot now. And it’s all about kind of matchmaking, finding the right ML talent to place on our client’s project. We hire these folks as Amii employees for a set term and give them a lot of support and mentorship, but really, they’re dedicated to work on the client’s project and move that forward. It’s a way for our clients to get access to talent without having to do the recruitment themselves. Amii has some pretty good brand recognition, we’re able to bring really great talent in and then place them on these industry projects.

A potential benefit of the system is the client having the opportunity to hire these folks after the term with us is done. We want this talent to stay here. We don’t want brain drain. We’re giving the client a bit of a leg up so that they can try the talent out, try out the project, get a feel for what machine learning actually is, what do we need to make it successful, and then ideally placing the talent within these companies in a longer term so that these companies really become AI companies and are able to move their own initiatives forward in the future.

How long is the term that they sign up for normally?

Generally, four to twelve months.

It’s something we figure out at the beginning, depending on the complexity of the project and how many problems we’re trying to solve. We find the longer, the better. Machine learning projects to do in four months can be challenging. There’s a lot more to it than just building ML models. Heavily reliant on the data that’s collected from the client that’s handed over to us, that helps us build the models. The longer we have, the better it is to iterate and cycle through all of the opportunities.

The work is experimental and exploratory in nature. Amii is a research institute; we can’t always guarantee the outcome. A longer runway just gives us more time to do that research and make sure that we’ve exhausted our options and pursued as many things as possible because it’s hard for us to say, “This is the method that’s going to work best.” You have to try it and see.

What are some examples of challenging business problems that your team has worked on with these companies?

I alluded to it, definitely data preparedness is a big challenge. Ongoing industry perception of data preparedness is different than what a machine learning scientist would think is ready for a machine learning model. And access. How easy is it for the client to hand over the data to us in a way that is consumable for our ML models. That’s why we do like longer projects because it gives our team time to work with our clients through those sorts of data preparedness challenges and set them up for success.

Garbage in is garbage out, if you hand us garbage data, we’re going to create a garbage model. We really need quality data. And there’s a little bit of a learning curve for clients. Industry perception, again, of what quality data is, what are the examples that we need to see to be able to predict things in the future. It’s just a literacy thing, making sure that we’re speaking the same language, they understand the limitations based off of whatever data they have access to when they understand what’s going to set us up for success.

You need examples of what you’re trying to predict in your dataset. If an event is really rare, it’s going to be hard for us to ever anticipate it happening. We could build a really accurate model of something that just say 99% of the time accurate because it’s never predicting the 1% time that something does occur. Again, just making sure that the client understands what we need to build accurate models.

We’ve seen even seemingly simple problems can be highly complex depending on their dataset. At the outset, having an initial discovery call with a client, we do have to anticipate the length of time that we will need. But sometimes when we start peeling back the layers of the onion, we realize, no, this is much more complex than we thought because of these data complexities.

Other challenges, lack of commitment from subject matter experts needed. When we partner with our industry partners, we really need them to continue to come to the table because they are the domain experts and usually the data experts too. We’re not like a dev shop where we can just take the data, build the model, and hand it over to them in the end. It’s very, very collaborative. And the more that our industry partners put in, the more that they’ll get out because they’ll be able to guide us in the right direction, make sure that the predictions that we’re making make sense to them from a business perspective, that we’re targeting the right metrics, we understand what success is for them.

We do need a multidisciplinary team around us to support the projects and it takes more than just one machine learning scientist to build a successful model that’s going to impact a business positively. There’s lots of challenges. Those are the ones that came to mind.

You personally believe that AI should be a force for good. What are some ways that you think AI can positively change the future?

The thing I like most about my job is we work with clients from across all industries, solving very different problems, but all of them are really being used for some sort of positive change. And Amii has our principled AI framework that ensures that we’re doing just that. From the contracting stage, we’re making sure that the projects that we’re working on with our industry partners are being used for that positive change in an ethical way. All the projects I get to see are being used for good and positively changing the future.

One thing that comes to mind, in Alberta more often than not now we’re dealing with wildfire situations in the summer. This year especially, even in April, it was bad. We recently partnered with Canada Wildfire. It’s a research group out of the University of Alberta. 40 years of weather data tied to severe wildfire events. Working with them to better predict these events in the future so we can better prepare the resources that might be needed, have the teams go in and temper the environments before it gets to a stage where the wildfires are raging. I think that’s just being in Edmonton, I don’t know if you were here last week, but it was very smoky.

When I arrived Sunday night (May 21, 2023) it was quite smoky.

It’s devastating. It ruins communities. It takes people’s homes away. Having to breathe particulate in the air isn’t great, but the devastation is very immense. That’s one interesting (project) that’s close to all of our hearts.

Another area we’re working in is the agriculture space. How are we going to feed our growing population? We’re working with the National Research Council on a protein abundance problem. Trying to make sure the plants that we’re growing have higher protein content to feed our growing population and using machine learning to be able to make those predictions.

Reducing emissions is another very popular one. Working with companies in the oil and gas sector to make sure that the processes and systems and tools that are used are as efficient as possible. We’re working with a water treatment plant out of Drayton Valley, which is a small town in Alberta, making sure that that water treatment plant is running as efficiently as possible and that we’re creating as much clean water for the community as possible. Precision medicine as well.

The list goes on. Literally, every company we work on its these sorts of projects, these sorts of causes. It’s hard for me to pick a favorite because when you think about it, they all have the possibility to have a incredibly positive impact on the future.

What is your vision for the future of AI or robotics?

My exposure to robotics has really been in the supply chain. It’s where robotics are already being used, but it’s also how do we enhance them with AI to build on existing systems and automation, again, through more efficient processes? The supply chain is obviously interested in increasing throughput, fulfilling more orders more quickly, and more efficient decision-making. On the robotics side of things, again, my exposure has been building on top of existing robots to make them smarter and better.

I think more generally, the future from what I see industry doing is still very human-centric. Robotics are used as a tool, as an augmentation to humans. Maybe robotics being deployed in conditions that are dangerous to humans where we shouldn’t be exposed to the environments. Robotics are a great replacement for us in that case to keep us safer. There’s also really cool research being done by our fellows and bionic limbs, so easier control and movement of people who do need that support. All very much still tied to humans and their use of these tools but making it easier for them to use and making their lives easier through these new systems.

In terms of the future of AI in general, this is just such an interesting time to be in this space. Industry is finally getting it that AI is here and it is going to change everything and you can either lead or be led. I think one of Amii’s visions is to have every company comfortable with the technology, aware of what it can and cannot do, and really willing to experiment and iterate on implementing it in their business to solve some of their toughest problems.

Up until now, I think maybe there was a perception that it was just tech companies that were AI and ML users, but now it’s becoming more apparent that ML can be deployed in essentially every organization. It’s not always the right answer, but there’s usually a use case for it. I’m hopeful that the future is companies becoming natural AI companies themselves by getting more literate and familiar with the technology and aware of how they can use it for their business.

Thank you for the amazing interview, readers who wish to learn more should visit the following resources: