From Cognitive Science to LLM Ranking Optimization: Jason Yang’s Entrepreneurial Journey
Mother of Success Interview · October 6, 2025
Opening
Xiao He:
Thank you so much, Jason, for joining us. Let’s start by talking about your background and what you’re building now. Could you briefly introduce yourself?
Jason Yang:
Sure. I did my undergraduate studies in Computer Science and Japanese. After graduation, I came to the U.S. for a master’s in Computer Science, focusing on Machine Learning (ML) and Natural Language Processing (NLP).
During my studies at Johns Hopkins University, I became interested in pursuing a PhD—specifically in modeling human cognition using neural networks. That led me to an interdisciplinary PhD program in Cognitive Science at Brown University. But during the program, I realized certain things and ultimately decided not to finish the PhD, instead graduating with a second master’s degree.
From Financial Data to Social Recommendation Systems
Jason Yang:
After graduation, I worked in New York as a Machine Learning Engineer—first in finance, mainly doing signal detection: identifying valuable signals from complex datasets, such as predicting market-moving trends in trading scenarios.
Later, I moved to a social media company, working on Search and Recommendation systems. In simple terms, personalized recommendation means that when you browse Facebook or Quora, the content you see isn’t ordered chronologically—it’s ranked by algorithms modeling your interests and interactions on the platform. Personalized search adds another layer: it also considers your search queries in ranking.
I see signal detection and search & recommendation as a continuous path. The former extracts meaningful signals from noisy multidimensional data; the latter presents those signals to users in a more efficient and relevant way.
Entrepreneurship: Bringing Recommendation Systems into the LLM Era
Jason Yang:
Now I’m building a startup that provides tools for companies and brands to improve their visibility and ranking in large language models (LLMs)—and to monitor their performance across these models.
On the surface, this problem—GEO (Generative Engine Optimization)—looks like a continuation of traditional SEO (Search Engine Optimization) in the LLM era, just on a new platform. Many competitors take this approach.
But for me, this problem is really an extension of personalized search and recommendation logic.
In traditional recommendation systems, we model the conditional probability between users and content—given a user A and a piece of content B, we predict the probability of positive interaction.
In LLM environments, this changes: platforms like OpenAI or Perplexity don’t share user data. You only know the content side (the brand), not the user side. The model sees X (brand info), but Y (user info) is missing.
My product attempts to reverse-engineer that missing Y: based on the brand’s characteristics, positioning, and target audience, we infer a precise Ideal Customer Persona (ICP). We then model the personal user journey through which different personas encounter LLM results—essentially reconstructing the missing user dimension. This transforms a complex problem into something tractable.
The company’s name, Egaki, means “to depict” or “to portray” in Japanese. The concept manifests in several ways:
I help you depict your ICP precisely.
For each segment, I design the most effective self-portrayal—so your ideal users encounter your best version in LLM search results.
I continuously portray your brand’s ranking performance across LLMs and provide improvement strategies.
This way, companies can optimize how they appear in the emerging AI search ecosystem.
Studying Japanese: Language as a Mirror of Its Time
Xiao He:
I noticed you also studied Japanese and even passed the N1 exam. What inspired that combination?
Jason Yang:
It came from a particular time and place. When I finished high school, Japan was still leading globally in AI, NLP, and especially robotics. I’m from Northeast China, close to Japan, where many Japanese companies invested, and cultural exchanges were common. My school even placed top students into Japanese classes, believing strong learners could pick up languages faster.
So it was partly chance, partly inevitability.
Xiao He:
Do you think your multilingual background has influenced your work, life, or entrepreneurship?
Jason Yang:
Quite deeply—but more internally. I’ve never studied or worked in Japan, but learning languages helps me see the world from different cultural perspectives.
Now that I’m building a startup in the U.S., understanding the market requires stepping outside emotional or cultural biases. You can’t just read Chinese or English media—you must learn to switch lenses.
Sometimes I read Japanese media to see how they report on the U.S. or Asia. Language is a window—it helps you perceive a multidimensional reality from new angles.
From Language to Cognitive Science: Academic Crossroads and Turning Points
Xiao He:
You later studied Cognitive Science at Brown University. How did that come about?
Jason Yang:
At the time, I believed NLP was one of AI’s most challenging areas. People called it “the crown jewel of AI”—if you can solve language, you can solve almost anything in AI.
I like challenges, so I chose NLP. Cognitive Science was an interdisciplinary program, bringing together students from computer science, psychology, neuroscience, and more. I wanted to move beyond the pure CS bubble—because I believed AI’s core questions can’t be solved by computer science alone.
Xiao He:
You mentioned later pivoting away from that path—why?
Jason Yang:
When I entered Cognitive Science, I realized it felt like “pre-Newtonian physics.” Researchers were running experiments, collecting behavioral data, and trying to connect brain structures to outcomes—but without a solid theoretical foundation.
I wasn’t comfortable with that kind of black-box inference—you see inputs and outputs but can’t explain what happens in between. When we don’t even fully understand the neural system of a lobster, attempting to reverse-engineer human cognition feels premature.
So I thought: instead of staying in speculative metaphysics, why not return to the empirical world? That’s when I decided to leave academia and use machine learning to solve real-world problems.
From “Speculation” to “Practice”: Engaging with Reality
Xiao He:
That’s quite philosophical. You’ve moved from the metaphysical to the physical world.
Jason Yang:
Exactly. I used to enjoy abstract thinking, but not anymore. To truly understand “cognition,” you must practice in reality.
I’ve read many books on philosophy and psychology, but I later realized theories detached from lived experience are like castles in the air. Everyone’s circumstances are shaped by their own personal journey, and no universal theory from a study room can prescribe human behavior—just as no stochastic process can perfectly model real-world decision-making.
I prefer the complexity and unpredictability of real life over abstract speculation.
From Academia to Finance: All-In
Xiao He:
Was it a big shock moving from Brown into the financial industry in New York?
Jason Yang:
Huge. You go from a world where publishing papers sustains you, to one where you must create real value.
In finance—or any real industry—you need “skin in the game.” You can’t complain about the rules while wanting to win. Either commit fully or step out.
That was when I truly began to grow. I learned the importance of first principles—to think about value from the ground up, replacing biased assumptions with real-world coordinates.
From Goldman Sachs to Quora: From the World of Money to the World of Knowledge
Xiao He:
Later, you left finance for Quora—a very idealistic environment. What drove that change?
Jason Yang:
I worked at Goldman Sachs and a hedge fund. That period combined tech and business well, but the data was closed, models were constrained by explainability requirements, and there was little direct user interaction.
I wanted to work somewhere with open data and real user feedback.
Quora was perfect: massive content ecosystem, real-time interaction data—ideal for search and recommendation research.
That experience completed the final piece of my puzzle before entrepreneurship. It gave me clear self-awareness, a sense of direction in both market and technology, and ultimately inspired my current startup idea. That’s why I say “signal detection → recommendation systems → brand ranking optimization” is a continuous spectrum.
The First Startup Step: From Supplements to Generative AI
Xiao He:
How did you find your current direction?
Jason Yang:
Initially, I tried a project that directly applied my past experience: a signal detection and recommendation system for the supplement industry.
But I quickly realized the industry was too traditional and closed—conferences cost thousands of dollars, and if you weren’t an insider, people ignored you.
More importantly, the market was too small (TAM too limited) to support a scalable startup.
So I pivoted to a much larger opportunity—optimizing brand visibility in the LLM ecosystem. Generative AI is a true “new frontier.” Major opportunities always arise at the edges of old systems. Few times in life do we witness such epochal shifts; being born in this era means we should board the ship and explore.
On Co-Founders: Not Because VCs Say So
Xiao He:
You’re currently a solo founder. Have you considered finding a co-founder?
Jason Yang:
I’ve thought about it, but I don’t believe in doing it for formality’s sake. Some VCs recommend it for risk control—so if you collapse, someone can take over (laughs).
But for me, alignment in vision is everything. When the direction is clear and the product compelling, the right people will naturally join.
So I’m open to collaboration, but won’t compromise just to meet external expectations.
Book Recommendation: The East India Company and the Seas of Asia
Xiao He:
Lastly, could you recommend a book or podcast to our readers?
Jason Yang:
Yes—one I love is The East India Company and the Seas of Asia (興亡の世界史 東インド会社とアジアの海).
It tells how European nations established East India Companies from the 16th to 19th centuries and adapted to different cultures. For example, their strategies in India, Africa, Japan, and China were completely different.
It reminds me that startups must also adapt to their environments—constantly negotiating, adjusting, and redefining rules. History isn’t just about the past; it’s a manual for survival and evolution.
Xiao He:
Thank you, Jason, for sharing your story. Wishing you all the best with your startup!
Jason Yang:
Thank you.