Ronit: Welcome to today's webinar building BreezoMeter’s unique technology and culture - our CTO tells all. Before we get started I wanted to go over a couple of housekeeping items. Today's session will be recorded and emailed to you afterwards, there will be a Q&A session at the end of the webinar. To ask questions please use the Q&A panel that can be found when you click on the Q&A button at the bottom of your zoom screen. For those of you joining us for the first time my name is Ronit Margulies and I will be your moderator on today's webinar.
On today's agenda:
And of course your questions.
Today's webinar will be hosted by our co-founder and CTO Emil Fischer. Emil has a BSc in software engineering from the Technion with years of experience in high-tech and major defense projects. Emil is an expert in software developing management methodologies and techniques, an entrepreneur at heart and at practice, and one of our founding fathers. If it's your first time joining us today BreezoMeter is a big data company that provides actionable personalized data for air quality, pollen, weather, and wildfires to businesses from numerous industries as well as governments. Take it away Emil.
Emil: Thank you for the warm welcome and for joining us. So actually I’m an intrapreneur by nature. I always wanted to invent new gadgets and software and change the world and this was my passion since I was a child. But let's get back to reality.
I think it's a good picture and it represents everything.
Ronit: Talk to us about this time, Emil.
Emil: So before establishing BreezoMeter, it was a different time. I was working in a place with a bad cultural environment where an employee was just a worker bee, where the words and ideas were often muted by force, by fear, and staying comfortable was the underlying conversation between workers. And I was supposed to get my tenure as well. I remembered that I felt stuck and I knew it wasn't enough and I felt like this is actually my last chance to be what I wanted to be. An entrepreneur. But still actually if I go back in time it was a hard decision for me to leave my company and establish BreezoMeter. I also think of that time when my wife was pregnant and I was about to get my tenure as I mentioned. We were looking for a house to buy and we had financial obligations like my wife who was in the middle of her second degree, so it was very hard. It was a very hard decision. But it was then Ran, my best friend in the last 21 years and today BreezoMeter’s CEO, approached me with an idea that answered actually all the key elements missing in my career path.
According to the World Health Organization, air pollution is a 21st-century epidemic and according to the EEA the European Environmental Agency, air pollution is currently the most important environmental risk to human health. The idea was to make environmental data like air pollution, but not only, visible. It was a project that I could believe in strongly. A space to be creative to build something significant for society. To do it from scratch and as I mentioned to invent things and change the world. That I would be able to establish a startup that puts focus on the right culture and celebrates its employees, and it would be their contribution and their culture that would be the secret sauce to the success of the company.
Ronit: I have to say, Emil, it's an ambitious goal. But I mean here we are years down the line and the dream was realized. And I can say actually personally as part of the team that I'm grateful that these were your strategies for success. But I mean our listeners, I don't think that they're going to buy that this was a straight line to success. We're talking about a startup that was taking on a big data task in a field that was only emerging and actually pretty dependent on how many consumers really cared about air pollution.
Emil: So there you mentioned there were many challenges.
Ronit: Just a few…
Emil: This is one of the things that makes BreezoMeter unique. I can name several challenges but there are so many. We need to raise awareness about, for example, air pollution, what it does, and how harmful it is. We need to educate the market and educate the people. Because air pollution is invisible, it's hard to fight an invisible enemy and sometimes you don't understand that the enemy is right next to you because it's invisible, but it shortens your life. We sell APIs, we sell data, we provide services. But if we focused on APIs, selling APIs is hard, and selling environmental data is even harder because it's a niche. We also provide our data to many different business segments such as pharmaceutical companies, the automotive industry, smart cities, IoT, smart home devices, and many many more, even cosmetics.
It's always hard to create products that speak in a new language for different business segments of what is the air quality or what is the pollen level at your location in real-time because it's really something unique, it's something that has never been done before. We also don't just provide data, we provide accurate data, so providing accurate data is a big challenge. It's not just garbage in, garbage out. It's really accurate data, personalized data, and really democratized data. It's your location. We also needed to deal with SLA, performance, big data problems, we'll talk about it soon. And of course, to provide more massive data you need a strong multidisciplinary team of software engineers, environmental engineers, and Big Data experts, and machine learning experts, and atmospheric scientists, algorithm specialists, and more. And they need to speak in one language, although each and every one of them studied started something differently. We need to develop many products and we also need to provide our products to consumers and businesses.
Ronit: Not an easy task. Essentially you're creating a possibility for people that don't know the possibility exists, people and businesses, within this multidisciplinary team and data. So I mean we're a technological company, is technology the answer here?
Emil: So many digital initiatives fail because of an overemphasis on technology, and I'm the CTO of the company saying that… But today technology for me is a commodity, and in most cases, startups don't fail because of technology barriers. Technology alone would not be the means to our success.
The biggest challenge for us and I believe should be for any successful company, is finding the best people. Recruiting A players for BreezoMeter and highlighting A players for us, because maybe there are other A players for other companies. I always like to quote Guy Kawasaki, who was the chief evangelist working with Steve Jobs twice before leaving Apple and after returning to Apple he said, always hire A players. Because if you will hire B players then they will hire C players and you will end up with the Z players that will hire bozos.
So we are not looking for mercenaries, we are looking for people to be part of our family. We're looking for people that need to be connected to our agenda of improving the well-being of billions of people worldwide who are exposed to environmental hazards every single day. They need to be connected to the agenda somehow. If it's, for example, due to the Freedom of Information level or if they want to do good or if they want to leave the world a better place for their children. So they need to be connected to our agenda.
Ronit: I can tell you for example when I joined, I was looking for a way to make a difference, and all these big ventures, these big ideas seemed too big for me to have an effect on the world. And this was something that really connected with me to hear I was able to affect change via the work that I would be doing at BreezoMeter.
Emil: Exactly. As I mentioned, we are not looking for mercenaries, we are looking for people that are passionate about what we're doing. People that their work is also their hobby, and for me, this is very important. we're looking for people that are team players because of the multidisciplinary team and multidisciplinary effort. We look for self-learners, people who are positive but yet are humble, and people who will go to battle with us when needed. People that I can trust to be there when they are needed. and also we don't like yes men. We want people to challenge us and ask difficult questions sometimes.
The team at BreezoMeter is extremely talented. They are being approached by the biggest and the most prestigious companies. So I believe we should put focus on the question: what makes an A player happy to stay at BreezoMeter? so every day again and again they will continue to choose to stay at BreezoMeter, because it should be a symbiotic relationship. For me, each employee needs to prove that he deserves to work at BreezoMeter, but also vice versa. BreezoMeter should make sure it deserves to hire such talents.
So if I focus on what makes an A player happy and stay, this is something that I thought about a lot in the last six years, and I shortened my list to the four following points. I really believe that if you work with people that are also A players they also teach you, they push you up, and it's a different working environment rather than just working with B players. I want other people to teach me so they need to be A players. Also, if people feel that they are doing something that is actually being part of something that is bigger than the employee, for example, you know that you have a global impact. Your mission or your vision is really changing the world, making the world a better place. So it will create a lot of motivation for employees.
Also, of course, I don't want people to be bored, so they need to be challenged in their domain. They need to see that their career is moving forward in their profession as expected. And also they need to feel that they are moving the needle. They are contributing significantly to the main effort and to the success of the company.
Of course, I also believe that managers are a crucial part of the happiness of employees, together with the culture of the company. Things like openness and being able to ask any question that you want. being able to challenge your peers, etc. This is crucial. This is also why we believe in our flat structure not in pyramids, like in other companies.
Ronit: So who are you in this cultural phenomenon?
Emil: I have a small role. It’s just to recruit the best people you need. People that are better than me and make them happy and stay. help them, show them where they need to aim. Remove obstacles and make sure that they are working on a meaningful task that will make a difference, of course. I don't want people to just work on their own things that in the end we’ll throw away.
Ronit: Okay, so, we have the team. The structure has been put into place for a culture to thrive and it thrives. But Emil, I mean at the end of the day the technology is still essential. I mean terabytes of data, the limited team, the limited time, the limited funds, the promise of performance, coverage accuracy, relevancy, stop me please. Talk to us about the technology please Mr. CTO.
Emil: You're right, Ronit. First of all, we use many raw data layers. Many. I can name some, for example, we take information from official governmental monitoring stations that measure the concentrations of different pollutant types. And the data is out there. we're relying heavily on the Freedom of Information Law, but it is scattered across the web. We collect data from hundreds of different data sources. And it's a headache. you need to scrape the data, find it, get legal permission to access the data, you need to build the scraper or the crawler to collect the data. So this is only monitoring stations.
We also take that data from traffic, because traffic, of course, affects the levels of air pollution. We take meteorological conditions, of course. You don't need to be an environmental engineer to understand that wind dramatically affects how air pollution is dispersed. We take satellite information and also we use smoke from wildfires in our fires layer. We care about the smoke that is coming from wildfires.
As opposed to other air quality data providers that just report raw data from monitoring stations, we use all of the aforementioned data layers as real-time input datasets that come into our algorithms to calculate government environmental data at almost half a billion geographical locations around the world. We provide personalized recommendations for the best action to take according to your personal sensitivity and the air pollution outside. To provide you with the air pollution level, the pollen levels at your location, we need to calculate the air pollution and the pollen levels around 500 million geographical locations around the world.
So yes. We also deal with massive data load, petabytes per month, it’s a lot of data even for big companies. We have to be able to calculate billions of pollutant concentrations around the globe using our complex algorithms. More than 35 algorithms. And we use thousands of virtual machines to calculate billions of pollutant concentrations and we analyze almost two terabytes of data each hour again and again. It's quite a lot. I think 10 years ago it wasn't possible even for big companies because it’s a lot of computing power and it costs a lot of money.
Ronit: Also, as you guys can see in the video that I'll start to play in just a moment, pollution changes. It just disperses over the course of minutes and hours, over the course of the day, from one location to another. so I'll just take a stab in the dark on this one, Emil, but that must need a lot of computing power all the while providing this data in real-time not to mention doing so in dozens of countries in parallel. The pressure on our developers must be massive.
Emil: Exactly. This process has to be focused and efficient and as a Big Data company, choosing the right cloud platform was extremely critical for us. After examining several solutions we chose to work with the GCP, Google cloud platform. It enables us to focus on delivering value to our customers rather than dealing with DevOps, maintenance, and scaling issues. GCP also provides us a technological peace of mind and enables us to focus on new technological efforts and our critical challenges to keep driving our business forward. If you want to I can share with you several examples of how exactly GCP does that?
Emil: Ok. So, we are big fans of managed solutions, because it saves a lot of time. We don't just want to work with metal or just virtual machines but rather, we prefer platform as a service rather than just infrastructure. so Google's managed solutions have been a game-changer for us. For example, thanks to App Engine, dealing with API queries is super dynamic. We don't have to worry about load balancing, operating systems, scanning issues, server maintenance, and erratic load changes. you just deploy your code and let Google handle everything else which is great especially for small teams or relatively small teams. Also, GCP is a sandbox for our data scientists but not only for our data scientists but also for our software engineers.
GCP data studio and data lab significantly reduce the cycle of developing and deploying new algorithms. So for example back in the day, our environmental engineers wrote code in MATLAB, and then software engineers had to shift that code into Python and then deploy the code to the operational environment. But actually today environmental engineers write code. Yeah, they write code in Python and they deploy their code to the cloud, which is really mind-blowing. If you had asked me this question six years ago I would tell you it’s a big challenge but now it's reality.
In our vision, when you take your son and daughter to the park, or you are on your way to work or to have dinner with your loved ones, you know what you're breathing and how it impacts your health. Therefore, for such a responsibility, we had to choose the most scalable, flexible, and innovative platform, and I must say that the fact that their team is located in Israel and we work very closely with them made it much easier so GCP was and still is a great match for us.
Ronit: I want to just sum it up for everyone. I sent them an email today asking if you guys had any questions and somebody asked them what is your biggest asset? How would you answer that question?
Emil: I think it won’t be a surprise when I say the biggest asset is not the virtual machines or the office or the equipment but just the people. We like the A players that I mentioned. They are the secret sauce for the success of the company, and all of our achievements are thanks to them.
Ronit: Amazing. Thank you so much, Emil.
We've come to the Q&A portion of the webinar. Please, everyone, feel free to send in your questions, we have time for a few. so I just want, as you guys are sending in your questions, I want to answer one of the questions that were sent in today from the email, how do you measure accuracy?
Emil: Okay, so we have several methods to measure accuracy. First of all, accuracy is a huge and significant part of our daily work. We were actually really fanatic about it. In our TV screens in our offices, there are accuracy metrics that are being shown to reflect the accuracy each hour again and again. We have a separate and isolated service that calculates our accuracy for the last hour. We really have hundreds of different metrics about accuracy. But regarding your question about the method, I can mention one method but there are others. We use the leave one out cross-validation method to check our accuracy. Let's assume that the world consists of tens of thousands of sensors or monitors.
Each monitor monitors the concentration of a different pollutant type. So let's talk about ozone. For example, what we do if we look at ozone, we take each station and remove it from our main algorithms as if the station wasn't there. We calculate the concentration of ozone at the location of the station. And then we compare it to the actual reading that the station reported retroactively after several hours because each station has also an inherited delay of reporting data. if for example, you will check four stations in the US you will see a gap between one to three hours of delay in reporting data.
So we do it each hour for tens of thousands of monitors around the world, so we have a really good understanding of our accuracy metrics because we do the leave one out methodology for each monitor, again and again, each hour. I hope I answered your question. By the way, another method is also a temporal method. What I described was the spacial method, how we measure our prediction in space. There are also methods to check our accuracy in time. As I mentioned, stations have a delay. In some countries, the delay can reach more than five hours. So we don't care about the concentration levels of what happened five hours ago because air pollution changes dramatically in space and time. I want to know what the concentration of each pollutant type is now. So we also use machine learning algorithms to predict what the level of each pollutant concentration is now. And then after several hours, we compare the readings that we thought versus the actual readings that came after several hours.
Ronit: Thanks, Emil. So, we are short on time. Regarding the questions you sent in, we will follow up with you tomorrow as well as share the recording. We talked about the technology, we talked about the culture, we showed you a bunch of pictures of our beautiful faces. I just want to say again thank you, Emil, for joining us. It was great, it was informative. Thank you guys for joining us for continuing to come back, and we will see you–
Emil: And thank you, Ronit, for hosting this webinar. But not only this webinar. Thank you for initiating this great effort of webinars that really provide more insights and information to people outside BreezoMeter.
Ronit: Come back to our next one guys. Have a good one.
Emil: Bye guys.