Hrant Davtyan, Founder and CEO at Pinsight
Episode 34
According to McKinsey Institute, AI could deliver an additional output of $13 trillion to the world economy by 2030, which would boost global GDP by nearly 1.2 percent a year. The future is big for AI.
In this episode, Hrant Davtyan, Founder, and CEO at Pinsight joins HyeTech Minds to talk about his startup journey. Hrant shares his thoughts on the importance of AI and Data in the decision-making process. Hrant also talks about the challenges and opportunities for Armenian startups to expand AI development.
In midst of global health crisis, Hrant has started Pinsight with a mission to give every decision-maker the power of making smart, insight-driven decisions fast. Pinsight is the first AI platform that allows product and customer experience teams to get AI-driven predictions and insights with a click of a button. Pinsight empowers company’s analysts and decision-makers with instant access to AI-powered predictions and insights.
Hrant’s Bio
Founded a Data science consulting agency Pinsight and scaled it to 40+ projects in 2 years
– Helped 10+ startups to create AI products
– Received Ph.D. combining Data Science with Economics
– Taught data science and business to 1000+ offline students
– Taught data science and analytics to 6000+ online students
– Consulted/Advised on 100+ AI/data science projects/products
– Helped UN (UNDP/UNFAO) as well as gov. bodies in implementing AI/data science projects
– Watched pretty much all Marvel and DC movies and TV series
Listen Now
Subscribe:Apple Podcast | Spotify | Google Play | Breaker | Pocket Casts
Interview Highlights
Narine:
Narine: Hi Hrant. Thanks for stopping by at HyeTech Minds. How are you doing today? How is it in Yerevan?
Hrant: Hi Narine. Thanks. I’m excited to join your podcast. I’m doing well. Yerevan is fine. The weather is a bit, it seems like it’s going to rain soon. How are you?
Narine: I’m doing great, thanks. As long as the weather is good, I’m fine.
So, Hrant, you have been in the data science world for many years. I’m really curious to learn how you got into the tech world. I wonder if you can first start by just telling us how you got into the tech world, data science world.
Hrant: Well, that’s an interesting story. So my background is fully in Economics. But in Economics, there is a subcomponent called Econometrics, which is fully data-driven and very quantitative. I always liked my econometrics courses. And when I was starting at the university, we had a TEA, who actually apart from TEA sessions was doing individual office hours on Data Science topics. I got excited to start to learn by myself. And that’s how I got into data science.
Narine: Wow, you’re a self-learner, super. I know you also hold a Ph.D. in Data Science. I’m curious to know how you went from academics to entrepreneurship. ’What sparked the idea to launch Metrics?
Hrant: When I was doing my Ph.D., my friends and I started a small Research Center, a nonprofit organization, where we wanted to just use Data Science for the public good, and we ran some projects.
Afterward, we saw that this is not really sustainable. So if you really want to use Data Science for the public good, the nonprofit option is not really sustainable, at least for the resources we have. Then, given the skills and expertise I have, I decided to basically test myself in the consulting business. By ending up having some consulting clients myself and seeing that my own time is not enough to actually serve all the clients, I decided to start the company. So basically, it was a conversion from consultancy, to company consultancy, from individual to company.
We started with just me and one of my students, previous students who had already graduated University at the time at Metric.
And then the idea was to basically come up with a very strong team of data science and machine learning engineers, which are not many in Armenia and in the world, as you know. So I was lucky to teach at the university exactly the same courses that I needed to basically do the consultancy projects that Metric. So I basically started hiring from the best of my previous and current students, and by any doubt forming a strong team of data scientists for Metric.
Narine: That’s super empowering when a professor recruits their own students. I think this should be an example for others. So talking about your team, can you take us inside your team? How big is it now?
Hrant: There are 21 people right now. So they are not so big, but not so small as well. So for Data Science, the Machine Learning team is actually quite big.
Narine: That’s a pretty big team for a new company. So, you founded Metric in 2019. What is Metric about? What are some of the problems you’re trying to solve?
Hrant: Our vision is to help companies make smart decisions fast, and we want to allow everybody to have instant access to insights from their own data. The main solution we provide is a product called insight, which basically helps companies to just connect their raw data to help insight what they’ve often lost and often wanted to get insight on, that can be used in marketing in customer analytics, as well as in product and politics. Typical use cases.
For example, let’s say you want to gauge user satisfaction analysis, you just connect your data and understand which users are actually not satisfied and who are satisfied and what are the key driving factors behind. That’s basically the main product Metric provides. But you also have some other services which include web First categorization. So basically, you provide millions of websites, we have to tell you there are business categories and industries. We provide document analysis, which helps you to extract valuable, better financial information from big documents. And so far.
Narine: I was going through your website and noticed that you also offer a Machine Learning training program. Can you talk about it?
Hrant: So we are very launched, basically, concentrated on trying to flourish the Data Science ecosystem in Armenia. So several times in a year, we are running free training programs, it’s like completely free of charge. And we don’t really have any motivation.
Apart from contributing to the ecosystem. We are running training programs regarding not only Machine Learning, but also general data analysis, like Excel, or Python, etc. We’re people, mostly students, and early career professionals can come on the level of their skills. But most importantly, they have something called data size Summer Internship Program, which actually they’re planning to launch soon, for 2021, which helps university students who have the knowledge but not the experience, and complementary gain experience and go to the job market, there are some of them, but not everybody. So there are many job markets, talented people who are ready to go to companies and solve problems, not just people who are talented, but also talented experienced.
Narine: That’s an amazing initiative for Armenia. That’s so important to provide hands-on experience to students, To prepare them for entry-level positions. This can put Armenia in a stronger position in the global tech scene.
Hrant: Yeah, certainly.
Narine: Hrant, we all understand that data is important. As some experts describe, nowadays data is an “oil.” But, from your expertise, how can you describe the importance of data for companies? Why do companies need to rely on Data Analytics in the decision-making process?
Hrant: It’s basically a matter of opinion, or belief versus facts. If companies are not relying on data for decision-making, they will quickly lose the competitive edge. Because in the competitive landscape, you cannot only rely on your own, on your, on your subjective opinions or subjective beliefs many times, because just biologically, our brains are biased. We have a lot of biases in our brain, because of neurobiology, we can come up with wrong conclusions. So it’s very important to actually rely on decisions also by analyzing data. And you cannot really get insights from your data if you don’t have data, right? So it’s very important to have the correct pipeline to collect data, which can many times give you interesting insights that can be very unexpected that otherwise, you wouldn’t notice.
Interested?
Do not miss out on latest startup news, tech trends, access to VCs, Accelerators, and tech jobs at the top U.S. companies.
Narine: I’m glad you talked about bias in the decision-making process. Many businesses and industries are accelerating AI adoption as the most effective way to make unbiased decisions. But, technology is not always accurate and we can not rely on AI 100%. How should companies use data analytics to reduce bias in their decision-making process?
Hrant: That’s a great question. So many times individuals are short-sighted and kind of make conclusions on so-called small samples, for example, they see that summer in their neighborhood, a certain product is selling well enough. And they conclude that this product is highly demanded. So that could be correct about your neighborhood, doesn’t mean that product is highly demanded by the general public.
If you make your business decisions only on the limited information that you observe yourself as a human being and make conclusions, let’s say on the information you get from neighborhoods, you can make your own decisions and go on and start to try to produce that product even more in areas that it’s completely on demand. So one bias that I typically observed in people is making decisions based on small samples. And in case I’m using data analytics, in case of having a strong data analyst and data science teams, you should probably be able to observe the real behavior that is experienced in your data, not just in your neighborhood.
Narine: As I said many industries today heavily rely on data analytics. In your practice, what industries, companies are more engaged with data analytics for the decision making process?
Hrant: So my experience shows that when it comes to customer analytics and marketing analytics, those two sectors are very heavily relying on dates nowadays. This is very good news because marketing can be divided into components of the right kind of romantic marketing, which is not really let’s move on to analytical marketing.
Many times you can see not only use cases, industry best practices driven by marketing analytics and marketing departments, so in my opinion are getting at customer teams are really driving most of the use cases in the data analytics world, but the other sectors like finance or even HR are growing rapidly, you can nowadays see a lot of HR tech companies that are trying to apply analytics in order to understand applicant behavior, employee behavior and make a conclusion based on that.
Narine: That’s an important topic HR. HR is one of the top industries that are noticeably accelerating AI adoption as the biased tool for workforce management. Companies use data analytics for all three stages in workforce management – recruiting, promotion, and employee engagement.
55 percent of U.S. human resources managers already said AI would be a regular part of their work within the next five years.So, how can HR be more effective in using AI for workforce management in a way that will make the decision-making process less biased?
Hrant: That’s a great question. So if you use AI, you can at the same time be more biased and less biased, you will be less biased. Because if you use AI and use data analytics, your conclusion will be driven by data and not your own subjective feelings or subjective opinions.
But at the same time, your AI models your data, analytical tools are as good as the data behind right. And the data is created by nobody other than humans. And many times humans are themselves. I actually remember a very interesting situation we had last summer with some of the big-name companies, for which the face recognition system was completely biased.
Narine: Amazon case, you mean?
Yes. Last summer, it was Amazon that stopped providing his recognition, called recognition, the face recognition system because it was terribly biased. So we had a similar situation with Microsoft actually, who took down a big, big textual corpus, because the textile notations are racist. So many times the data can itself be a bias that if you use an AI model, the results can also be biased, which means it will get rid of the individual bias objectives, emotions, feelings that you can have, for example, in HR use cases, but at the same time, you have to be cautious not to introduce a different type of bias, something like ethical or racism, or gender bias or anything else that could potentially be observed data, because AI is learning the patterns you have in data if your data is biased, automatically a written contract as well.
Narine: You know one of the conversations recently massively discussed will AI, robots replace humans? World Economic Forum’s AI is expected to replace 85 million jobs worldwide by 2025? What is your take on how far AI will impact the workforce?
Hrant: That’s actually a good question. So several years ago, I was participating in a conference. It was about Economists. But there is a field in economics, which is called labor economics, that observes how the labor market is changing.
So I met one Nobel laureate in labor economics, who told me about his working paper that was trying to understand how AI will affect the labor market. What he was saying is that if he split individuals into three groups, low skilled, medium-skilled, and highly skilled individuals, then medium-skilled individuals will be affected by AI most reasoning behind was that so he was doing all mathematics behind, but the intuitive reasoning was that the low skilled people will have a competitive edge in terms of price.
Compared to AI, the highly skilled people will have a competitive edge compared to AI in terms of quality, so that the middle-skilled people, don’t have the competitive edge in terms of quality and not in terms of price as well. So they will have the most trouble. So if you’re asking me, where are they I will replace the human or not, I think, yes, it will, for some component, but at the same time, it will introduce a lot of new jobs. For example, because of computer vision, there are a lot of human tasks that are not really done by humans anymore. Computer Vision is often difficult to cross regarding images and videos. But at the same time, in order to have this good computer vision algorithm, unit humans will label and annotate images, so it is taking some jobs, but at the same time, it is creating new jobs, which in this example is human and features.
So yes, I think it will take some jobs, but it will certainly create new ones as well.
Narine: That’s an interesting insight. Good to know this. Another problem is data privacy. Amount of data companies accumulate is stunning. Time to time you hear stories that big companies, like LinkedIn ,yahoo had data leakage. What you think about this?
Hrant: Well, that’s actually something really problematic. And that’s exactly why we have regulations like GDPR, etc. And I actually learned today, I don’t know how accurate it is because it feels like a rumor that I learned today that the European Union is going to introduce a big amount of fines if AI is used in certain use cases that they consider that it shouldn’t be used.
So I think the way to actually make sure that AI is used for good reasons is to have strong but also well-sought regulations. I mean, strong is not enough, it should be well discussed with industry professionals. And also the regulations that will not really restrict the growth of the industry, but at the same time will restrict the bad so-called use cases of data, which can either introduce fake content or introduce privacy issues.
Narine: Europe has started to be more serious about data privacy. I’m all for regulations. I’m all in for more regulations.
Hrant, talking about the data impact on decision-making, sometimes we forget to understand what are the limits of decision-making by an algorithm?
Hrant: The first and foremost limit is data. Many times when I’m telling people that I have a product that can provide you sides and recommendations, automatically, they think that it’s going to also provide some general business recommendations, like hire a new person, or change your business strategy.
AI is limited by the data that you have, meaning that it can only provide insights into the limits of the data. And as a result, the largest challenge that I observe now is something called causality.
So causality is something that statisticians and econometricians have studied for many years. And this is a hot topic in AI. What AI is doing is finding patterns, right, like quarterly. Patients, but if it finds patterns, it doesn’t mean that there is a causality, there’s a pattern between A and B, it doesn’t mean as causing B, right? So the biggest challenge in using AI for decision-making is that many people just use patterns followed by a claim that it is causality, which is actually a big problem because otherwise, you’re going to make wrong decisions based on that.
Narine: So what we are witnessing today is a deep impact of AI on the economy, and humans. According to McKinsey Institute, AI could deliver an additional output of $13 trillion to the world economy by 2030, which would boost global GDP by nearly 1.2 percent a year. The future is big for AI. In this sense, from your perspective what are your thoughts on the future of AI in the decision-making process?
Hrant: That’s a great question. So I am actually for automation. So currently, AI is built by human experts, people like me, data scientists, and machine learning engineers. But unfortunately, we don’t really have enough supply of data specialists in the world, there is a big shortage gap.
I mean, according to statistics, 70% of companies are not able to hire data scientists and they have problems in the process. And as I do, the compensation is very high, so many companies cannot really afford it.
So I see the future of AI. In automation, I see AI creating AI, we actually already have the first steps. There’s a technology called Automental, which is actually the technology behind our product as well, which creates machine learning models itself using machine learning. So AI is creating AI. And I think that’s the future to basically come up with a solution that will cover the demand gap in the market.
Narine: AI is really becoming huge not only for business, but also for government, countries. Today, you see global superpowers like China and the USA are competing for AI dominance in the world. Now I’m thinking about countries like Armenia, which are very small and need to be exposed to the international tech landscape. What do you think are some of the challenges for Armenian startups to expand AI development?
Hrant: So the main obstacle that I currently see is that companies are really interested in hiring data sciences and applying data for decision making, which is very good, but they want to hire senior people.
And we don’t have enough senior people in the market. So what we really need to do is to have a lot of opportunities for junior people, a lot of internship programs, which unfortunately are not so many. Last year when we announced our internship programs. I have only seen another internship program in the market for this year. I have seen the third one? So I think this will become more like the standard, do you offer an internship program by leading data companies in Armenia? Otherwise, we cannot really go over the senior people because there are not so many people in Armenia who are seniors. So we have to give the opportunity for more young talents to get the experience and the good all of them inside our companies and become seniors themselves.
Narine: I understand they want to hire senior engineers. But, the question is that those entry-level engineers need the opportunity to get hands-on experience and grow experience. No one becomes senior at once. What do you think, how can the Armenian tech Diaspora be helpful in this?
Hrant: That’s a good point.
So I think there’s a lot of things that a Diaspora can do. I myself have met Armenian data scientists abroad, who maybe some of them migrated from Armenia. Some of them left Armenia to study data science abroad because we didn’t have courses in the universities of the time and then stayed there. And some of them are just not from Armenia themselves, they were born outside.
So I think we have a lot of smart talent in the Diaspora who can really make a lot of positive knowledge investment in the Data Science sector. We already experienced something like that this year. We were running a program called AI incubation program, together with Hero House. What we have is basically, data science experts, mentors, people who are very experienced in data science, from abroad, mentoring Armenia, university students to conduct research, and those people were really excited to do that. And at the end, when we had the feedback of students, it was really positive. They learned a lot of hands-on experience from top companies in Europe and the US or the UK, how scientists used to do it was very motivational. And I think it should become standard practice to kind of try moving the Diaspora and the Armenians abroad in the process of flourishing given the size and general technical system in Armenia mentorship is just one example. But I think there are many, many other cases.
Narine: I mean Armenians have always been known as talented scientists. And with the help of the right mentorship programs and tools, Armenia has a huge potential to expand its AI ecosystem. Continuing this conversation, what do you think we need to change in Armenia in order to attract more investments from abroad?
Hrant: Oh, that’s a very difficult question.
Many things we need to change really. Recently, the High Tech Ministry in Armenia was conducting a survey to understand whether the legal aspect is an obstacle for investment.
We don’t have things like employee stock options that you can give to your team if you don’t have typical investment agreements that startups have, like safe, etc. Those things are problematic for startups to actually legally be organized in Armenia.
So as you know, many companies are actually going out organizing themselves. In the law, they’re using clarity and stripe and similar tools.
So as a result of the investments, monitoring, from a monetary perspective, going to these US companies, so one of the things we have to improve on is basically the legal system first, and then the image of Armenia as a safe country where you can actually have a monetary investment from non monetary perspective, we have to be able to try. The Diaspora isn’t really motivated, it’s meaningless. People come to Armenia, let’s say teach a class at the university and basically, students on 10th class systems are not the trade-offs, people will be getting motivated, right?
So they have to make sure that there is an atmosphere where these people can come and really feel comfortable and motivated to invest not only monetary investment to invest their time of knowledge to basically help the younger generation and existing confidence in Armenia.
Narine: Great point. This is something to explore furtherly. So, what would be the best way to learn more about Pinsight?
Hrant: So we are actually planning to start a podcast ourselves. So that’s one approach. But generally, we have our website, y’all our Facebook page, where we are constantly providing our updates on our performance, our products https://www.pinsight.ai/
Soon, we’ll be more active actually on social space, so they can really learn from us on Facebook or websites. But they can also feel free to reach out to me on LinkedIn and ask questions or discuss any topic that they find interesting and that I can contribute.
Narine: Thank you so much Hrant for your insights. Good luck with all your initiatives and stay safe.
Hrant: That was a great experience, super exciting to talk to you now. Thanks a lot for this podcast and for organizing this.
We are Armenians & We are innovators…… Stay up to date on the Armenian startup scene and the latest tech trends, get access to the Armenian-founded VC and Angel Investors, connect with peers Join our community today
Links to the Founder and everything you want to know about Pinsight
- Founder’s LinkedIn https://www.linkedin.com/in/hrantdavtyan/
- Website: https://www.pinsight.ai/
- Facebook: https://www.facebook.com/PinsightAI
- LinkedIn https://www.linkedin.com/company/pinsightai/
- Email: hrant@pinsight.ai
Links to HyeTech Minds Community
- Host’s LinkedIn https://www.linkedin.com/in/narine-emdjian/
- Website: https://hyetechminds.com/
- Facebook: https://www.facebook.com/hyetechminds
- LinkedIn: https://www.linkedin.com/company/hyetechminds/
- Twitter: https://twitter.com/hyetechminds
- Instagram: https://www.instagram.com/hyetechminds/
Hye Tech Minds 