The Eclipse Founder and Former Citadel Quantitative Researcher Is Empowering Others With His Knowledge
SAN FRANCISCO – Neel Somani has spent much of his career working inside systems most people never see, let alone understand. His work has spanned electricity markets, blockchain infrastructure and machine learning models, fields defined by complexity and constant change. The connective thread across each of them is a deep curiosity about how things actually work.
“I tend to follow curiosity more than trends,” Somani said when we sat down for an interview about his life and career. “I want to understand problems from first principles.”
From the outset, it is clear that Somani believes knowledge is a form of leverage. Over the years, he cites mentors and teachers who shaped his thinking, and today he is focused on paying that influence forward by sharing what he has learned.
Scroll through his recent videos online and a different side of the San Francisco-based founder comes into view. In January 2026, Somani began sharing short educational clips on TikTok explaining power markets, pricing mechanisms and the hidden logic behind energy systems. Several videos quickly gained traction, circulating among students, engineers and finance professionals drawn to their clarity and lack of jargon.
“I was a quant researcher on Citadel’s commodities team covering power, or electricity, and I’m going to teach you enough about power that you’re going to be smarter than 99% of people that you talk to,” he said in one video, speaking directly to viewers.
He explains how central authorities called ISOs or RTOs gather bids from suppliers and forecast demand before calculating what he describes as the “economically efficient price.”
“It’s kind of complicated how they come to that price,” he said, “but you can just know that it is the efficient price. There’s no funny business going on.”
He walks through dispatch order in plain language. “You can kind of think about it as the system takes the cheapest power to go first. So that’s like solar, wind. Then you have nat gas and coal and everything else.”
He then distills the concept into a single insight. “The price that everyone pays is the amount of money it costs to produce that very last megawatt of power.”
At the end of one clip, he signs off simply. “I hope that this gets you started in your power pricing journey.”
The videos reveal something essential about Somani. He does not simplify because he has to. He simplifies because he wants to.
Neel Somani is the founder of Eclipse Labs, Ethereum’s first Layer 2 powered by the Solana Virtual Machine. Since launching the company in 2022, the project has raised $65 million, establishing him as one of the more technically rigorous founders working at the intersection of blockchain infrastructure and applied computer science.
Despite those credentials, Somani repeatedly returns to a simple idea: knowledge becomes more valuable when it is shared.
Inside the Berkeley Experience That Shaped Somani’s Approach to Hard Problems
Neel Somani has long been drawn to systems that operate at massive scale and require precise reasoning. From electricity markets to distributed networks, he gravitates toward environments where complexity is not a barrier but an invitation.
“The power market in the United States is solved via a general optimization technique called mixed integer programming,” he said, describing the optimization models that determine electricity pricing. The intellectual challenge drew him to Citadel’s commodities desk, where he applied quantitative modeling to real-world markets and gained a deeper appreciation for how theoretical mathematics translates into practical decision-making.
That experience reinforced a pattern that would define his career: follow the hardest problems.
Long before Citadel, that mindset took root at the University of California, Berkeley, where Somani pursued a rare triple major in mathematics, computer science and business administration. He graduated with a 4.0 GPA, earning recognition as a Haas Undergraduate Scholar and receiving the Cal Alumni Association Leadership Award, distinctions that placed him among the top students in one of the country’s most competitive academic environments.
At Berkeley, he immersed himself in a culture that valued intellectual rigor and interdisciplinary thinking. Participation in the EECS Honors Program and Phi Beta Kappa helped deepen his technical foundation while reinforcing his interest in connecting theory to real-world systems.
His interest in formal methods began during this period, shaped by coursework emphasizing first-principles reasoning. “Formal methods are cool because they’re derived from first principles, and they have profound implications that extend beyond computer science,” he said. “Theoretical computer science is more practical than you might expect. Every undergraduate computer science student learns about the halting problem. It reveals that there are limits to algorithmic reasoning.”
Working in Professor Dawn Song’s lab exposed him to applying formal reasoning to privacy and security. “That was the first serious project that I worked on in formal methods,” he said, recalling research focused on proving that machine learning systems satisfied formal definitions of privacy.
After graduating, Somani worked as a software engineer at Airbnb before moving to Citadel. The transition reinforced his belief that the most interesting work sits at the intersection of theory and application.
After his time in quantitative finance, he turned toward blockchain infrastructure, where he saw a clear scalability gap. Ethereum’s security made it dominant, but throughput limitations created friction for developers building high-performance applications. Somani founded Eclipse to address that challenge, combining Ethereum’s security model with the performance advantages of the Solana Virtual Machine.
The company quickly gained traction, positioning Eclipse as a serious player in next-generation blockchain infrastructure.
“I’ve always been drawn to problems where the surface looks simple but the underlying system is incredibly complex,” Somani said.
Opening the Black Box of Artificial Intelligence
In recent years, Somani has focused increasingly on artificial intelligence, particularly the challenge of understanding how modern models behave internally.
“Right now, safety and interpretability in machine learning is preparadigmatic,” he said. “There isn’t any established way to certify that a system is safe, or that we fully understand it.”
He believes formal methods offer the strongest path forward. “Formal methods are the gold standard because they’re the only way to establish strong, principled guarantees about programs,” he said. “But we’re a long way off from being able to apply them to machine learning systems.”
Somani points to robustness as a core challenge. “You don’t want a model where if you change the inputs by a little, then the outputs change wildly,” he said. “That would imply that the output is unreliable and unstable.”
He also highlights reliability at the infrastructure level. “A tiny bug in how you wrote the GPU code can lead to hidden errors that are hard to unearth,” he said.
His project Symbolic Circuit Distillation aims to bring more rigor to interpretability research.
“When someone analyzes a model, they might have a guess as to what it’s doing under the hood,” he said. “But there’s no way to really prove or disprove that hypothesis.”
Knowledge as a Form of Leverage
For all the technical depth of his work, Neel Somani’s motivations remain grounded in a clear philosophy: understanding complex systems is only valuable if that understanding can be shared and applied.
He believes the next phase of artificial intelligence will require stronger guarantees around reliability and safety as these systems become more embedded in everyday decision-making.
“In an ideal world, for high-stakes or mission-critical machine learning systems, the entire workflow would be formally specified at all levels,” he said. “As machine learning models write a larger percentage of our code, we might expect more out of them.”
Looking ahead, Somani sees a shift underway in how progress is measured. “The field is moving from building bigger models to building better understood models, and that shift will define the next phase of AI,” he said.
He also anticipates breakthroughs in how systems retain and use information. “AI today has an extremely limited context window,” he said. “Over the next few years, we’ll see breakthroughs that allow AI systems to ingest entire codebases.”
Beyond research, he supports higher education through a personal scholarship program and continues to share knowledge publicly through videos, blog posts and open datasets.
He believes understanding itself is a form of leverage, and that broader access to knowledge leads to better systems and more thoughtful innovation.
“The long-term opportunity is not just making AI more powerful, it’s making it more understandable, because understanding is what ultimately builds trust,” he said.
Near the end of our conversation, Somani was asked about the direction of his growing social media presence. He paused briefly before answering, explaining that the motivation is less about strategy and more about curiosity and sharing.
“I’ve been fortunate to learn from people who took the time to explain things to me,” Somani shared with a smile, “and I think passing that forward is one of the most meaningful parts of the work.”