Maria Greicer Podcast Transcript
Maria Greicer joins host Brian Thomas on The Digital Executive Podcast.
Welcome to Coruzant Technologies, home of The Digital Executive Podcast.
Brian Thomas: Welcome to The Digital Executive. Today’s guest is Maria Greicer. Maria Greicer is a seasoned executive with over 18 years of experience in AI driven technology startups across Europe, USA, and Israel, specializing in machine vision AI applications. Combining an entrepreneurial mindset with a medical background, Maria brings a unique perspective to the AI industry.
She holds a BA from Reichmann University, specializing in entrepreneurial management and information technologies, and is deeply passionate about emerging technologies. AI advancements and continuous learning. Maria has led strategic growth initiatives and multiple leadership roles, including CEO, VP of Partnerships and other executive positions.
Currently Maria is the Vice President of partnerships at Keymakr, working with enterprise clients on complex technical solutions for machine vision AI. She creates and optimizes training data to ensure high quality data sets for AI model development.
Well, good afternoon, Maria. Welcome to the show!
Maria Greicer: Glad to be here.
Brian Thomas: Awesome. I appreciate you making the time. I’ve been excited, looking forward to talking with you. I know you’re hailing out of Calgary right now. It’s awesome. Very cold. It’s cold here. So, it must be very cold up there. I actually had some friends in high school from Calgary.
So pretty cool. So, Maria, I’m going to jump right into your first question. If you don’t mind in having worked with AI driven technology startups across Europe, USA, and Israel. What are the key differences you observed in the A. I. Ecosystems of these regions and how have these experiences shaped your approach to innovation?
Maria Greicer: Okay. So that’s a very good and very interesting question. It’s like a lot to share here. So, it starts from a little bit of a background. So I grew up in Israel. So, most of my work experience before moving to Canada, before moving to us and also living sometime in EU is, was in Israel. So I’ve been working in startups since I was.
For like 18 years, I’ve been working in technology startups and there is a very, I would say significant difference in those three regions, how technology is developed and how well AI specifically is developed, but technology as a, like a whole theme. And I would say the main difference. Is the approach and the risk appetite of people and companies.
So, for example, if we take Israel. So, in Israel, there is a lot of risk appetite and basically, I’m not, I’m not exaggerating almost every person who works in tech. He wants to open a startup because there’s always like friends of friends who open the startup. So, opening a startup, creating a startup, developing new technology.
It’s not something that’s so uncommon or a big deal. So, people used to it. So therefore the mindset is, okay, we have to think of new technology. Okay. We have to create a new company. Okay. To do that, we have to do something that’s unique, different and crazy. In a way. So, we have to take risks. And the approach is that, okay, we’ll take those risks.
We’ll invent this idea. We’ll try. We’ll give it a try. If it works, great. If it doesn’t work, also great. We’ll try something else. And this is like sort of like a machine that’s going on. So, people try to open a company. They succeed, great. They sell it, they open the next one. If they don’t succeed, also great.
They close it and open the next one. So there’s, I would say more turnaround in trials, like what technologies people try to develop, but also like the more you try, right? Like the more you succeed. So the percentage of AI technologies in general, but also AI technologies that come out of Israel is pretty high.
I would say. Thanks to this, uh, unique approach because just people willing to take more risks. It’s not something that is frowned upon or not welcomed. I’m looking at U. S. So I was, I was living in U. S. in Washington, D. C. for a while. Uh, that was an interesting experience. It’s not the tech capital of U. S.
of course. There are a number of technology companies in Washington. There is like interesting developments there. What I noticed about U. S. like United States. Yes, they tend to take risks, but because there’s so much capital, there’s so many things to do. There are so many options. So yes, there’s of course, a lot of technologies come out of us as well.
But if you kind of compare it to the amount of risk and the amount of. Willingness to do crazy innovative things, I would say, again, I can’t compare per capita, not in total, because of course, US population is way, way, way bigger. But in the US, it’s, I would say, a little bit lower. It’s a bit more conservative in a way.
And if we’re looking at, uh, comparing to Europe, so in Europe, again, it’s, uh, also different approach, different people. The main value that I found that is important, or not one of the values they found is important for companies is stability, stability and profit and predictability. Which is great in general, but it’s not great when you try to innovate.
So, it’s basically from what I’ve noticed, there is less tendency, less willingness to do riskier moves or to do things that are completely don’t make sense, but could lead to incredible breakthroughs or could lead to incredible new technologies developed just because again, like the. The willingness to take risks and the encouragement of the society by the society to take risks is a bit lower.
Like it’s not really encouraged as much as in US, especially in Silicon Valley or in Israel.
Brian Thomas: Maria, thank you so much for sharing that. I have to say, I’ve, I’ve done probably several dozen podcasts in Israel, and they say Tel Aviv is really the Silicon Valley of Israel, if not the world. You obviously are proof of that.
And you talked about some of the innovations and the Israelis are very innovative and very persistent in the startup community. So, I appreciate you sharing that. That’s just amazing. And so, Marie, I’m switching to the second question here. At Keymakr, you focus on creating and optimizing training data for machine vision AI.
Could you elaborate on the processes and standards you employ to ensure the data sets meet the high quality requirements for AI model development?
Maria Greicer: Yes, of course. So, uh, Keymakr is an Israeli based company, but we operate globally. And what we do, our core focus is ensuring ground truth and high quality training data for machine vision.
So, our niche specialty is machine vision AI. We don’t do anything else. So this is actually can be very interesting and very tricky. Like what is machine vision AI and how does this data look like? So, it’s basically visual data that helps to train machine vision applications, like any camera based application to see and recognize what’s going on around, recognize the environment, recognize, uh, like anything from, uh, medical applications, like.
Tumors on MRI scans to self-driving cars, capabilities of recognizing the lanes, the traffic lights and pedestrians and other cars on the road. So it’s very, very broad applications of machine vision AI. And so what. The way we approach this last task, so in a way, like Keymakr is industry agnostic, but we have some industries that are more popular than others, like automotive or agriculture or security or medical.
But in a way, our approach is applicable for any, any industry out there. And our approaches, I would say we’re like, in a way, like Elon Musk, we do everything in house. So, our system, we develop our own proprietary system for creating the training data, Keylabs AI. It’s an annotation platform. It’s a powerful annotation platform that supports all types of annotations.
That’s our proprietary technology, our IP. So we use this platform for creating the training data. Now, in addition to the platform, we need human input, which is Annotation in a way from scratch or validation of models output. So here, the key is also everything we do is in house. Our employees work from our company computers.
Those are professional operators. All the data that we process goes through four levels of quality control, which are custom tailored for every project. Now, I would, I don’t want to go into much to the technical details here, but I would say the key here is this, first of all, that we have full control over the operation of the training data preparation.
Second is that we tailor our QA. Approach and we tailor our project management approach to every different project, regardless from what industry it’s coming from. So, this combination allows us to produce high quality training data sets, regardless of complexity of the project.
Brian Thomas: I appreciate that. Ray, you obviously have a lot of strong knowledge and background in this, and I appreciate you taking us through that, the process there, there’s so many use cases for this, and I’ve had so many people that worked in, for example, the autonomous vehicle space, and there’s just so many applications beyond that, but I really appreciate you highlighting some of the key things there that you do at Keymakr.
Maria, given your passion for emerging technologies and AI advancements, what current domains and machine vision AI do you find most promising, and how is Keymakr positioning itself to leverage these developments?
Maria Greicer: So, I would say number one domain is autonomous driving, autonomous driving, autonomous vehicles.
Now this is both relevant for driving itself, for environment recognition, road recognition, like the actual autonomous driving capabilities of the vehicle. There’s so much work done in this domain right now by companies from startups. The large enterprises all over the world, like everybody is trying to tackle this challenge and develop systems that would allow fully autonomous driving.
So every company they do this similar thing, but they’re. Approaches are different, so there’s enormous amount of advancements like every day in creating the system. So self-driving capabilities is definitely something that we’re going to see as a mainstream very soon. It would most likely, it would first take over locations that are men-less.
Like warehouses, some closed factories where we need to drive around, but there are no people involved. Basically, the less people are in the area, the more likely this facility would be fully autonomous. Because the danger comes from having people suddenly being in front of the machine. So, I would say autonomous driving, autonomous vehicles, that’s something that we definitely see taking place, being a mainstream technology pretty soon.
Another interesting thing about autonomous driving, so it’s not always about the vehicle, it’s also about the passenger. About what’s going on inside the cabin. So for most of the new vehicles that are coming right now to the market, the cameras are not only facing outside, they’re also facing inside. So the passengers are being, the passenger, the driver are being monitored.
So why is it, like, it sounds scary, like why the car is watching me, right? Like who wants their car to watch. But it’s actually very important and contributes a lot to the safety of the passengers. So, systems like recognition of distracted driving, or if the driver is tired, or for example, if it’s a mom and with her kids on the back seat and she’s turning back to see what’s going on and not paying attention to the road.
So, the systems. The systems are learning to understand the behavior and the state of the person inside the car and address or take control accordingly. So for example, if the person is tired and the system recognizes that the person is tired, it would take over the wheel in a way and prevent an accident.
Same thing with distracted driving. So really our vehicles are getting more and more smart, regardless of autonomous capabilities, but also like understanding what’s happening inside, like what’s happening to the driver and the passengers, which I know like some people find scary, but that’s, that’s where we are heading.
That’s where, where we are right now. Another thing to add here that is very interesting is agriculture. So recently, I would say over last two or three years, there is a lot of investment in agricultural related AI, both for managing crop growth as well as disease recognition. And the disease recognition of the plants, this is something that farmers So growers used to do more manually, which takes more labor and not always as accurate or precise.
And right now we have very, very smart systems that are taking over the disease recognition and the crop management, and this should increase the yields. This should decrease the waste of chemicals on the pesticides and really contribute to better food, less expenses involved.
Brian Thomas: Thank you. I appreciate that.
You covered quite a bit. I know that Keymakr has a unique approach to this AI development for machine vision. Obviously, autonomous vehicles is a big deal. You’re not only looking at starting off in more closed warehouse environments, but you’re looking to look at what’s happening inside a vehicle, which is cool.
Of course, you switched gears to agriculture. So again, appreciate you highlighting that. And Maria, last question, if you could briefly share. Most international enterprises are currently engaged in the development of AI systems. What are the unique aspects and challenges of working with companies like this, where there are very strict data privacy and data protection regulations, and you know, they vary from country to country for the same company.
Maria Greicer: That’s a very interesting question. And I’ll tell a little story that happened a few times. Actually, it happens like every month almost. So, the story goes like this. We have our R&D department we’re working with. We’re doing project estimations. We’re coming up with a project of a training data set that we’re going to create and annotate for this R& D department.
They need the data right now to hit their development deadlines. Everything looks great. We get a confirmation and then the project gets stuck in the legal department because the legal department cannot allow any risk associated with data acquisition. And then sometimes projects are paused or canceled or completely reshaped because.
The R&D team that’s located, for example, in US cannot use the data in the same way as they would regularly because the clients of the company are also in EU and the data protection policies in the EU are completely different. It’s a very common use case. It’s a challenge right now for many international companies.
And the way it’s addressed, addressed properly is first of all, of course, like regulation is very important. It’s important that we have data protections rules in place and kind of the more time passes, the more strict they get, which is a good thing in my personal opinion, but it’s definitely makes training data preparation and model development a bit more challenging.
So how do you source this data? How do you ensure that that is ethical, that is legal and you can actually Like the company can safely use the data, considering the different data protection, privacy protection requirements from different countries that they have to comply with all of them together. So one of the approaches that we have, especially for data creation.
Is when the data is collected, the data, wherever that is collected from each person or each individual in this case, whose data is being used, which is for facial recognition applications, for example, Oregon, whatever the application is, has to have a personal consent in place. So, all the data has to have some sort of proof that, yes, it was collected legally.
Yes, it was, there is an agreement from those people or companies or environments that it’s okay to use the data. There is a proof of this agreement and the data is processed in a secure way. What is secure way? It means it’s not shared with the public. It’s only used for the purpose that it was collected for.
So basically the data cannot be like shared or published or shown to any other in any other places. So here, what helps, uh, what we see is very important is the security, of course, the consent, but afterward, the next step is the secure processing of the data, especially during the annotation part. So that’s, again, one of the reasons we do everything in house.
So, all that is safely stored and, and securely processed. So, there’s no risk of the data being leaking outside of the organizations. And it was less of an issue, I would say a couple years ago. Right now it’s like number one question and most that all of our client ask. It’s like, okay, what’s the security?
How do you protect the security of the data? How do you ensure there are no data leaks, so can here have full control and everything done in-house? Not crowdsourcing, very important, not crowdsourcing, not outsourcing. Let’s say our personal key differentiator and benefit that we can offer.
Brian Thomas: Thank you. I appreciate you sharing that.
I know that data privacy and data protection has only gotten stricter. And of course, as you know, GDPR out of Europe and then here in the United States, California leads the pack as always when it comes to laws, right? But no, really do appreciate that. I think that’s so important. And you and I share the same sentiment about privacy.
I think it’s a good thing. We just need also keep ethics in the AI development, as you know. So, I appreciate that. And Maria, I just want to let you know, it was such a pleasure having you on today, and I look forward to speaking with you real soon.
Maria Greicer: Thank you so much for having me. It was a pleasure.
Brian Thomas: Bye for now.
Maria Greicer Podcast Transcript. Listen to the audio on the guest’s Podcast Page.
