AI in Real Life: With Dr. Teddy Lazebnik

Ryan Penn

Hello and welcome to AI Voices, the podcast where we explore the human and AI side of artificial intelligence combined.

Ryan Penn

I'm your host, Ryan Penn. Today, we have a very special guest joining us – Dr. Teddy Lazebnik. Teddy is a researcher and serial entrepreneur in the field of AI, with over a decade of experience building AI solutions in multiple areas, from finance to healthcare. He holds a PhD in Biomathematics and has co-authored over 80 academic papers on topics like personalized medicine and computer simulation. He's currently an Assistant Professor, where he focuses on using math and AI to solve biomedical and socio-economic problems. But Teddy isn't just about academia – he's also co-founded several AI outsource companies and delivered over 85 successful AI projects worldwide. Teddy bridges cutting-edge AI research with real-world impact, and we're thrilled to have him on the show. Today we'll talk about his journey into AI, his key contributions in making AI useful in healthcare and beyond, and get his thoughts on where AI is headed. And we'll do it in a way everyone can understand – no PhD required!

Ryan Penn

Teddy, welcome to the show! It's great to have you here.

Dr. Teddy Lazebnik

Hello Ryan, thank you for having me, it is a delight to be here today.

Ryan Penn

Before we dive into everything, could you share a bit about yourself, your background, and what you're currently doing?

Dr. Teddy Lazebnik

For sure Ryan. Nowadays, I am combining two fronts – academia and industry. In the academia, I try to develop and use new mathematical and computation models to solve problems in various disciplines. I would be happy to name you one specific area but I am in love with the fundamental mathematical nature different areas like Economics and Biology share so for me all these research subjects are cool stories to invent and solve mathematical problems.

Dr. Teddy Lazebnik

Let me give you an example, together with a lot of amazing collaborators, I am working on the usage of AI in personalization of cancer treatment. Amazingly enough, on a completely different project about informal economy mitigation, we were able to use the same math we used to personalized cancer treatment to reduce informal economy. If this is not considered modern magic, I do not what can satisfy this definition.

Dr. Teddy Lazebnik

In a complementary manner, in the industry, my team and I are working with Israeli and EU-based companies, usually A round funded startups, on their core algorithmic problems, giving them strategic algorithm and AI advantage in their businesses. While we provided services in chatbots and agentic AI, like many others, we mainly focus on deep tech projects that more often than not require real research to solve rather than just good engineering.

Ryan Penn

That’s a fascinating perspective—especially the connection between math and real-world problems. It seems you really found a sweet spot.

Ryan Penn

One question I always love asking my guests is: In your own words, what exactly is AI?

Dr. Teddy Lazebnik

It is an amazing question Ryan… Well, while I do not have a good answer myself yet, I really like to cite my PhD advisor’s definition for AI: “AI is everything your computer isn't able to do”. You see, we are currently at the third AI winter, it means we already have survived two winters – a heroic dead if you ask John Snow. However, to be less sarcastic, AI is not computational or mathematical concept, if you ask me, it is actually an economic and social property. After all, AI changes the economy and therefore the social norms while other algorithms are indifferent to AI advantages.

Ryan Penn

I like that definition—‘AI is everything your computer isn't yet able to do.’ It's a refreshing way of looking at it.

Ryan Penn

There's a lot of buzz about AI right now—some say it's revolutionary, others say it's just a fancy version of a Google search. What's your take—is AI genuinely groundbreaking, or just hype?

Dr. Teddy Lazebnik

You see Ryan, the question as you propose it does not capture the complexity in the field of AI itself. If you refer to AI as the data-driven revolution where we do not program algorithms directly but program generic algorithms that takes data and “learn” specific algorithms – I think it is easy to see that AI is indeed groundbreaking. On the other hand, let us be mindful about what AI is at the moment – a really fancy statistics engine. Currently, it is not proven to be able  to think, to be creative, or to even generalize very well for new problems and tasks. It is not to say I do not find AI amazing every single morning, but I guess we all need to be somewhat more critical about the objective state of AI and its meaning.

Ryan Penn

Interesting! It’s good to keep that realistic perspective in mind amidst all the hype.

Ryan Penn

You started making waves academically and professionally at a relatively young age. Could you tell us how your journey into AI began?

Dr. Teddy Lazebnik

I do not think my journey was into AI but rather to mathematics. An ongoing family story is that while three-four year old kids in the neighborhood played with a ball outside, I prefer to solve magic square tasks that my uncle gave me. From a young age, when my classmates wanted to be policeman and firefighters, my answer to whatever I want to do when I will grow up is to be a mathematician. I started my academic studies at a relatively young age, studying mathematics. However, during my bachelor’s degree, I noticed I really struggle with the highly abstract mathematics and actually enjoy the applied courses, which considered in close circles to be inferior. At the time I also worked as a programmer and found a real magic in the computer’s ability to solve math better than I can after I program the right instruction on how to do so. I guess after a few years doing that, and with huge luck of being around in the industry at 2016-2017, AI was a natural way forward. At the time I did not know that AI will become so mainstream but I enjoyed it so much that it was not so important.

Ryan Penn

Clearly, math runs deep in your blood.

Ryan Penn

Was there a specific moment when you realized, “Yes, AI is the field I want to dedicate myself to”? What inspired that moment?

Dr. Teddy Lazebnik

I am not sure I have a specific Eureka moment. In a more general sense, let me even guess that most of your guests will not have it. AI is relatively new for most and from the stories of my peers, people get into AI from different subjects such as math, physics and other fields in computer science – it's just a cool field. That said, of course, the fact it pays did not harm the decision to pick AI as my main field.

Ryan Penn

That's a practical answer—I guess sometimes careers evolve naturally without one specific 'Aha!' moment.

Ryan Penn

You have quite a diverse educational background. How did your studies shape your approach to AI?

Dr. Teddy Lazebnik

Coming from a mathematically rigorous discipline, I naturally gravitate towards modeling AI systems as structured, explainable processes rather than just black-box solutions. I also guess that due to my interest in biological systems, I commonly take natural ideas into my AI solutions, such as genetic algorithms, multi-agent optimization strategies etc...

Ryan Penn

Having that mathematical foundation clearly shapes your unique perspective on AI.

Ryan Penn

A lot of your work has focused on AI in healthcare and personalized medicine. Could you explain, in a simple way, what personalized medicine means and how AI helps doctors?

Dr. Teddy Lazebnik

Personalized medicine is about tailoring medical treatment to each patient rather than using a "one-size-fits-all" approach. Instead of giving everyone with the same disease the same drug or treatment, doctors look at a person’s unique characteristics—like their genetics, lifestyle, and medical history—to find the best approach for them. Now, when I am speaking of “doctors” I mean I can totally see AI-driven doctors performing large portion of diagnostics, monitoring, and treatment design in the near future. Even today, AI plays a big role in making this possible. It helps doctors analyze huge amounts of data—like genetic tests, medical records, and even wearable device readings—to identify patterns that humans might miss.

Ryan Penn

That’s really helpful—personalized medicine definitely sounds like the future of healthcare.

Ryan Penn

You’ve worked on models predicting patient outcomes, like complications after surgery. How does AI actually make these predictions?

Dr. Teddy Lazebnik

You see, Ryan, different clinical settings call for different AI solutions, so I cannot fully answer this question for all cases. However, I guess that one of the most interesting use cases is this: Given a patient and a single treatment decision, it is crucial to predict what that decision will lead to. Let me give you an example, a doctor deciding whether to prescribe a certain medication or recommend surgery, AI can analyze past patient cases with similar conditions and predict potential outcomes—such as the likelihood of recovery, side effects, or complications. This is often done using causal inference models or machine learning models. In practice, AI models can take in data about the patient—like their age, gender, socio-demographic profile, genetic profile, prior treatments, and lab results—and simulate different scenarios. They might predict, for instance, that one treatment has a 90% success rate but a higher risk of side effects, while another option is less effective but safer. This helps doctors and patients make informed decisions tailored to the individual’s unique health profile.

Ryan Penn

It's clear why hospitals would find this extremely valuable.

Ryan Penn

Ultimately, the goal is to move from broad, generalized treatment guidelines to highly personalized medicine, where every decision is backed by data-driven insights. You mentioned a concept called ‘decision trees’ in your work. Could you explain what a decision tree is, and why doctors like to use them?

Dr. Teddy Lazebnik

I am delighted to learn you read my recent studies. A decision tree is a simple but sometime powerful model. Its basic idea is that you have a question in head, this is usually called the “root” of the tree. In order to ask this question, for example if we expect complications for a given patient following some operation. At each point in the decision making, you refer to a question in the decision tree, usually refer to as a decision node. If you answer “yes”, you  move left and for “no” to the right. At each decision node the question is different – for example, is the patient male or female? You repeat this left-right moving several times until you get to the final node, a leaf node, which gives you an answer – yes or no. Of course, more complex versions of the example I just present are used but the underline idea is identical. I believe doctors like them very much because they are both explainable models and as they very closely follow the thinking process professionals, in general, and doctors, in particular, intuitively follow as revealed by many psychological studies. I would like to state Ryan that while decision trees are great, they are more often than not too simplistic and we need to educate clinicians to trust statistical validation and use less explainable models which are able to provide more accurate predictions compared to decision trees. Please do not get me wrong, there are many of these already deployed and we are on the right track but much more progress is needed.

Ryan Penn

It's like having a clear roadmap for making tough decisions.

Ryan Penn

You’ve also done research involving epidemic modeling, especially during COVID-19. How does AI or computer modeling help us manage pandemics?

Dr. Teddy Lazebnik

Yes, my interest in pandemic spread started in March of 2020, I wanted to make sense in all the numbers I saw on TV. So, I started to learn about the subject and learn about simple yet brilliant mathematical model from 1918 called S-I-R.  I recall thinking that it is too simplistic and indeed further investigation revealed that for over a century it has been improved. Nowadays, it is also combined with AI-driven modeling which supercharge its capabilities. Based on this mathematical foundations, many tried to tackle COVID-19. Getting back to your question, the straightforward answer will take us too much time. However, if I have 30 seconds, the main point I would like your viewers to know is that these models are useless in accurate pandemic spread prediction. It just the way it is…. There are many technical reasons to why it is like that but this is the bottom line. Nonetheless, AI to pandemic mitigation works amazing. We can test millions of virtual scenarios and policies to find the best one. The secret? We do not need to know if tomorrow we will have 100 or 110 infected individuals – we just need the order of magnitude and the direction of the dynamics and these models are amazing in predicting these.

Ryan Penn

Interesting—I never thought about pandemic modeling this way.

Ryan Penn

What's the biggest challenge when using AI for something as critical as healthcare or pandemic response?

Dr. Teddy Lazebnik

The biggest challenge when using AI in healthcare or pandemic response is making sure the AI’s predictions are accurate, explainable, and unbiased before doctors or policymakers rely on them. AI models are only as good as the data they are trained on. If the data is incomplete, outdated, or biased (e.g., underrepresenting certain demographics), the AI can make wrong or unfair recommendations, which could lead to disparities in healthcare. In addition, AI models trained on past data may not adapt well to new virus strains, emerging diseases, or changing healthcare practices. This was a major challenge during COVID-19, where AI models built on early outbreak data often failed to predict later waves accurately. As such, organization need to have a fundamental change in mind that AI-driven models are not silver bullet that once it is deployed it is working identically over time.

Ryan Penn

Accuracy and adaptability really are critical here.

Ryan Penn

I discovered you also worked on a fascinating project involving cats and their facial expressions! Can you share with us what this project was all about?

Dr. Teddy Lazebnik

Yes, indeed, it is a really fun research project. The project was a collaboration with Dr. Florkiewicz and Prof’ Zamansky. We used videos of cats from a cat café that capture their interactions. We wanted to investigate something called rapid facial mimicry (RFM). Simply put, micro expressions that are rooted in a response to a communication, like we human have. Scientists have known for decades that people and other social mammals such as dogs, horses, and orangutans perform RFM as a crucial part of social bonding. So, in this study, we trained a computer vision AI model on hours of videos of cat interactions. The AI used 48 “landmarks”—digital dots placed virtually on strategic places on cats’ faces—to capture 26 unique facial movements. These, in various combinations, create the hundreds of facial expressions cats make. Next, the researchers examined how these facial expressions changed when two cats in close proximity looked at each other. About 22% of the time, the felines mirrored each other. The mirrored expressions were subtle, sometimes just a modest flattening of the ears paired with a small wrinkle of the nose or a tiny raising of the upper lip. But when they happened, the cats began a friendly interaction—playing together, grooming each other, or walking together—almost 60% of the time. The AI was vital to making the discovery, as the facial mimicry is practically impossible to spot via the naked human eye, even among cat experts.

Ryan Penn

Wow, who knew cat facial expressions were so complex?

Ryan Penn

Beyond academia, you've co-founded startups and delivered dozens of AI projects. What motivated you to step into the entrepreneurial side of AI?

Dr. Teddy Lazebnik

What motivated me to step into the entrepreneurial side of AI was the realization that real impact happens beyond research papers. While I love academia, I found that many powerful AI ideas remain trapped in theoretical models or experimental settings, never making it into the hands of the people who need them the most. Starting my own ventures allowed me to take AI out of the lab and solve real-world problems—whether in healthcare, fintech, or decision-making systems. The challenge of building AI solutions that are not just innovative but also scalable, reliable, and useful fascinated me. Another big motivation was ownership and speed. In startups, you don’t have to wait years for grant approvals or peer reviews—you identify a problem, build a solution, and iterate quickly based on real-world feedback. That level of agility is something I found deeply rewarding. Finally, let us to be honest, making a good living is always a main theme when I am looking for my next adventure.

Ryan Penn

Taking technology out of the lab and into practice is inspiring.

Ryan Penn

Can you give us an example of one AI project you delivered for a business and the kind of impact it had?

Dr. Teddy Lazebnik

Of course, one of the key challenges in healthcare operations is , which is critical for optimizing scheduling and resource management. Traditional pre-surgery estimates often fall short because they rely on failing to account for variations in To address this, we developed a that enhances both Our approach leverages incorporating factors like Using an, we improved pre-surgery duration predictions by . Beyond static predictions, our system , analyzing real-time data streams. This dynamic adjustment led to an additional for surgeries lasting over three hours. By integrating these AI-driven insights, hospitals can reducing delays and improving efficiency in surgical departments.

Ryan Penn

That’s impressive—it shows how impactful AI can be.

Ryan Penn

As someone who bridges academia and industry, what's one common misconception businesses have about AI?

Dr. Teddy Lazebnik

One of the biggest misconceptions businesses have about AI is that it’s a plug-and-play solution—that you can simply integrate an AI model, and it will instantly deliver perfect results. In reality, AI is highly dependent on data quality, domain knowledge, and continuous iteration. Many companies assume that once they build an AI system, the job is done. But AI models require constant monitoring, retraining, and adaptation because data distributions shift, user behaviors change, and real-world conditions evolve. Without proper maintenance, even the best models can degrade in performance over time.

Ryan Penn

That's important—AI is definitely not plug-and-play.

Ryan Penn

For listeners who might want to get into AI, what advice would you give them about where to start?

Dr. Teddy Lazebnik

For anyone looking to get into AI, my biggest advice is to start with a strong foundation in mathematics and programming—especially in linear algebra, probability, statistics, and Python. AI is not just about using pre-built models; understanding the fundamentals will help you adapt to new advancements and build solutions from the ground up. I see too many “AI experts” that just import ready models and call them AI. If you really want to understand AI, you first need to understand the basics.

Ryan Penn

Solid advice for beginners—fundamentals truly matter.

Ryan Penn

What’s one key skill that you think will become increasingly valuable in an AI-driven world, even for non-technical people?

Dr. Teddy Lazebnik

One key skill that will become increasingly valuable in an AI-driven world is domain expertise combined with AI fluency. As AI becomes more powerful, its real impact will depend on how well it’s applied to specific industries—whether healthcare, finance, law, or education. People who deeply understand their field and can collaborate with AI systems to enhance decision-making will have a strong advantage. For example, doctors who understand how AI assists in diagnostics, or financial analysts who can leverage AI-driven forecasts, will outperform those who rely solely on traditional methods. This doesn’t mean becoming an AI engineer, but rather developing the ability to work alongside AI, interpret its outputs, and integrate it effectively into real-world workflows. The future belongs to those who can combine human intuition with AI-driven insights, making informed, ethical, and strategic decisions.

Ryan Penn

Combining human insight with AI seems like the winning formula.

Ryan Penn

If you could go back and give young Teddy one piece of advice when you were just starting your career, what would it be?

Dr. Teddy Lazebnik

Be more a people’s person. That is the short answer. The longer answer is that while technical expertise is crucial, relationships, collaboration, and communication matter just as much—if not more. Early in my career, I focused heavily on mastering the technical side, being the best mathematician I can be, the most techy person in the room, the one that has the answers. However, as I mature slowly, I learn that alone can go fast and probably some distance but in a good team, we can go the real distance, and my job is to make sure that the phase is not much slower.

Ryan Penn

Great insight—i think a lot of us can relate to that.

Ryan Penn

Outside of your busy academic and entrepreneurial life, what do you like to do to relax or recharge?

Dr. Teddy Lazebnik

Well, I really like woodworking, especially animal models and shooting from traditional bow and arrow – these activities clear the mind. As a geek, I also, of course, like video games and unfortunately have too deep of an affection for a good pint of beer.

Ryan Penn

Sounds like you have a great balance!

Ryan Penn

Do you have a favorite book, movie, or series about AI that you think everyone should check out?

Dr. Teddy Lazebnik

Yes, I have! Somewhat advance, but I really recommend the reinforcement learning course of UCL with DeepMind – it has the right mix of theoretical background and applied parts.

Ryan Penn

Thanks for sharing—it sounds like a great resource.

Ryan Penn

Teddy, this has been a wonderful conversation. We've covered so much ground – from your personal journey to technical projects. I think our listeners will really appreciate how accessible and insightful you've made these complex topics.

Ryan Penn

Before we wrap up, how can people follow your work or learn more about you? Do you have an online presence you'd like to share?

Dr. Teddy Lazebnik

Absolutely! People can follow my work through , where I share insights on AI, research, and other stuff. I also regularly post updates on about the latest trends in AI, healthcare, and entrepreneurship.

Ryan Penn

Thank you to all our listeners for tuning in to this episode of AI Voices. We hope you found it as enlightening as we did. If you enjoyed this conversation, please subscribe and leave us a comment.

Ryan Penn

Until next time, stay curious and take care!Goodbye everyone.