S 01 | Ep 49 Balancing Act

Immigrant Experiences & Entrepreneurship

Dmitry Shevelenko is the Chief Business Officer at Perplexity AI, which has raised $165 million in funding and is valued at over $1 billion. Its investors include big names like Jeff Bezos, Nvidia, Databricks and others. Previously, Dmitri served as a co-founder and President at Tortoise. He gained experience by working at various leading mobility and future of work companies: Uber, LinkedIn and Meta.

Key Takeaways

(00:00-09:05) From Facebook to Perplexity

(09:05-17:47) Perplexity's Mission to Save Time and Enhance User Experience

(17:47-24:55) The Evolution from Links to Answers

(24:55-34:33) How Perplexity AI is Changing the Game

(34:33-43:29) Navigating the Tech Frontier

(52:48-1:02:58) The Immigrant Hustle

📱 Apple Podcasts | 🎧 Spotify | 🔗 iHeart

Building the
Future of Search

Driven professional with a passion for building user-centric products, demonstrated through a career journey spanning roles at Facebook, Pulse, LinkedIn, Uber, and currently at Perplexity, where I contribute to revolutionizing information access through AI-powered search and natural language processing. 1 My experience encompasses product development, understanding user behavior, and leveraging cutting-edge technologies like LLMs to create innovative solutions that prioritize speed, accuracy, and readability, ultimately enhancing user experience and disrupting traditional search paradigms

1. From Facebook to Perplexity

Dmitry discusses his career journey, emphasizing a key principle: working on products he genuinely loves. He highlights how his passion for these products has guided him through roles at Facebook, Pulse, LinkedIn, Uber, and now Perplexity. Dmitry believes that if he enjoys using a product, others will too, which is a strong indicator of the product's potential success.

The conversation delves into why Dmitry finds the knowledge and answers provided by Perplexity so compelling. He draws a parallel to Ask Jeeves, an early attempt at natural language search, which failed due to technological limitations. However, Perplexity succeeded by launching when GPT-3 became available, allowing for effective summarization of web content. This timing was crucial, as it meant that Perplexity could offer a user experience that directly meets modern expectations for natural language answers.

Alex highlights that Perplexity has created a new category in the market, providing a unique user experience distinct from Google's. This distinct experience makes Perplexity's position defensible, as users now recognize it for specific needs that other platforms might not fulfill as effectively. Dmitry agrees, noting that while established companies like Google have their advantages, they also face challenges in shifting user expectations. By meeting these new expectations, Perplexity has carved out a significant niche in the AI-driven information economy.

I think a lot of startup success or failure boils down to timing. Perplexity launched right when GPT-3 became available as an API. The ability to succinctly summarize web content just did not exist before that. So if you had tried to launch something with Perplexity six months before, it would have been too early. If you had tried to launch Perplexity six months later, potentially too late. Because we and others were already going down, creating data flywheels and refining the user experience. (Dmitry Shevelenko)

2. Perplexity's Mission to Save Time and Enhance User Experience

Dmitry explains that startups like Perplexity benefit from their ability to focus solely on the user without the burden of defending an existing business model. This allows them to prioritize user needs and make quick, iterative improvements. The mantra at Perplexity is to get 1% better every day, emphasizing the importance of shipping updates regularly and maintaining a sense of urgency.

The discussion highlights Dmitry's experience at previous companies like Uber, where user expectations were met with clarity and control. Similarly, Perplexity aims to meet and exceed user expectations by providing precise, quick, and readable answers. When a query is entered into Perplexity, it is broken down into multiple sub-queries. Perplexity's search engine and web index then find reputable sources, extract useful information, and use advanced technologies like LLMs (Large Language Models) for summarization. This entire process happens in under two seconds, balancing speed, accuracy, and readability.

Alex and Dmitry discuss the technical aspects that differentiate Perplexity from other search engines. Perplexity focuses on delivering a seamless user experience by ensuring the perceived latency is minimal. This is achieved by loading sources first, which builds credibility and provides users with immediate content to evaluate. By streaming tokens and presenting sources before the full answer is ready, Perplexity keeps users engaged and informed.

Dmitry's background in anthropology and interest in psychology inform his approach to creating a user-friendly experience. He understands that perceived latency is as important as actual speed, and the user interface must be designed to keep users satisfied while the data is being processed.

There are three product priorities that are evergreen: accuracy, speed, and readability. Those are always in tension with each other. You can optimize for speed. It may come at the expense of accuracy and readability. You can optimize for accuracy. It comes at the expense of speed. You can optimize for readability. It may come at the expense of both. (Dmitry Shevelenko)

3. The Evolution from Links to Answers

Dmitry elaborates on the classic innovator's dilemma, explaining that startups like Perplexity have the advantage of focusing entirely on the user's needs without the burden of maintaining an existing business model. This allows them to make decisions that prioritize user experience and enable them to move quickly, constantly improving their product.

Perplexity breaks down user queries into multiple sub-queries, retrieves information from trustworthy sources, and uses advanced technologies to summarize and present answers in natural language—all within two seconds. This process aims to balance speed, accuracy, and readability, ensuring users get reliable information quickly and clearly.

The brothers discuss how Perplexity's approach mirrors the concept of putting a mirror in front of an elevator to make waiting more bearable. By providing immediate context through visible sources, Perplexity reduces perceived wait times and builds trust with users. This method not only enhances user experience but also allows users to evaluate the credibility of the answers they receive.

Dmitry explains that while not all users click through to the sources, knowing the sources are available adds value. It reassures users that they can find more information if needed and often inspires deeper engagement. Perplexity's design encourages follow-up questions, promoting a more in-depth exploration of topics and creating a richer learning experience.

The conversation also touches on the psychology of modern knowledge consumers. Perplexity aims to shift from a link-based economy to an answer-based economy, where users get precise answers quickly, similar to how billionaires optimize their time. This approach caters to users' desire for efficiency and depth in their search for information.

Finally, Dmitry notes that Perplexity's follow-up feature for pro users helps refine queries in real-time, providing more accurate answers. This iterative process of Q&A leads to a deeper understanding and greater satisfaction for users. Perplexity's goal is to create a delightful, easy, and efficient experience, encouraging users to engage more with the platform and explore topics thoroughly.

A good answer leads to more questions. What was interesting about the engagement point of view is people want to go down rabbit holes of learning. So oftentimes, you may end up spending more time on Perplexity than you would on Google. But not because you're having to wade through a lot of junk to get useful information. It's because you get a much deeper knowledge. You're inspired to keep probing and getting closer and closer to the ground truth that you're looking for. (Dmitry Shevelenko)

4. How Perplexity AI is Changing the Game

Dmitry emphasizes that Perplexity taps into the core human emotion of curiosity. By making it easier for users to explore their natural curiosity, Perplexity provides a fulfilling and engaging experience. The platform's design encourages follow-up questions, promoting deeper engagement and a richer learning experience.

The conversation shifts to the business aspects of Perplexity, highlighting its recent successful fundraising round, which attracted high-profile investors like NVIDIA, Jeff Bezos, and Naval Ravikant. Dmitry attributes this success to the product's transparency and the universal need for efficient information retrieval. Investors and users alike can easily evaluate Perplexity's value based on their own experiences with the product.

Perplexity has also received high praise from the media, with the Wall Street Journal ranking it as the top AI chatbot, ahead of major competitors like ChatGPT and Google's Gemini. Dmitry credits this recognition to the company's relentless focus on user experience and rapid execution. He stresses that maintaining this edge requires constant improvement and avoiding complacency.

Alex and Dmitry discuss Perplexity's strategic partnerships with telecommunications companies and phone carriers. These partnerships help expand Perplexity's reach and integrate it more deeply into users' daily lives. Despite being a relatively young company, Perplexity has managed to secure these collaborations by demonstrating its value and potential to these historically slow-moving organizations.

I think the core human emotion that we tap into is curiosity. If you can make it easier for people to activate their natural curiosity, they enjoy it and it's fulfilling for them. It's useful for them. (Dmitry Shevelenko)

Check the episode's Transcript (AI-generated) HERE.