Building the Future of Crypto Infrastructure: Leo Fan's Journey to Scaling Zero-Knowledge Proofs

When Leo Fan came to the United States for his PhD at Cornell, he was leaving behind a stellar academic career in China. But his path wasn’t just about traditional computer science—he was also a competitive gamer, nearly going pro in esports while playing Counterstrike and Warcraft. That blend of technical rigor and competitive spirit would shape how he approaches building products today.

These days, Leo is the founder and CEO of Cysic, a company tackling one of crypto’s biggest infrastructure challenges: accelerating zero-knowledge (ZK) proof generation. And his story is a masterclass in how technical founders navigate the messy reality of building hardware-software hybrids in emerging markets.

The Eureka Moment

During his engineering days, Leo discovered something powerful: a cryptographic system with remarkable properties that could fundamentally change how blockchain systems scale. Specifically, he became fascinated with zero-knowledge proofs—a way to prove something is true without revealing the information itself.

The implications are huge. Imagine transferring 10 units of cryptocurrency to someone without revealing your total balance. With zero-knowledge proofs, you can prove you have enough funds without exposing your wallet’s contents. Compare that to Bitcoin or Ethereum, where every transaction detail is public by default.

But there’s another critical property: succinctness. Traditional proof generation is computationally expensive and time-consuming. Zero-knowledge proofs, however, are compact and fast. You can verify a proof on a smartphone in 15 milliseconds, and the proof itself is only a few hundred kilobytes. This matters because you can compress an entire Ethereum block—containing thousands of transactions—into a single ZK proof. Instead of re-executing every transaction to verify a block’s validity, you just verify the proof. It’s a game-changer for blockchain scalability.

That’s when Leo realized: I want to build the infrastructure that powers this technology at scale.

The MVP That Validated the Market

Here’s where Leo’s approach gets interesting. For him, an MVP isn’t about showing off technical prowess—it’s about delivering actual value to paying customers.

The initial challenge was brutal: generating a single ZK proof using CPU took about two hours. That’s completely unacceptable when new Ethereum blocks arrive every 10-15 seconds. Leo’s team needed to dramatically reduce that timeline.

They built a GPU-accelerated version of their proof generation system. Using 14.9 GPUs, they managed to reduce the generation time from two hours to around 15 minutes. That was the breakthrough moment. When they showed this to potential customers, people were ready to pay for access to the service.

This is the critical insight: they weren’t selling hardware or software in isolation—they were selling a service. They built a blockchain to incentivize people to contribute idle GPU power to the Cysic network, creating a distributed compute infrastructure that could scale with demand.

The Pivot Nobody Talks About

Here’s where the story gets real. Cysic’s original vision involved designing specialized chips optimized for ZK proof generation. Leo’s team spent over a year on this effort. They had engineering resources allocated, customer conversations happening, and a clear technical roadmap.

Then the market told them something different.

As GPU and FPGA hardware matured, they realized these off-the-shelf solutions were “good enough” to handle current demand. Building custom silicon would require years of additional development, significant capital expenditure, and—most critically—uncertain market demand. The algorithms for ZK proofs were still evolving. Potential customers were hesitant to commit to specialized hardware when the standards hadn’t settled.

So Leo made a hard call: pivot away from the custom chip strategy and double down on software optimization and the distributed compute network.

This wasn’t a small decision. It meant restructuring the team, redirecting engineering efforts, and accepting that some of their early work would become sunk costs. It also meant deviating from the original long-term vision of what Cysic would become.

But it was the right call. By focusing on what the market actually needed now, rather than what they thought it might need in three years, Cysic stayed lean and customer-focused.

Reading the Market and Expanding the Vision

Leo’s approach to product roadmapping is refreshingly pragmatic: monitor the market, identify trends, and follow real customer demand.

When the Ethereum merge happened, it didn’t immediately create the computational demand Cysic expected. So rather than forcing a single narrative, they expanded their vision. They started supporting other computation workloads beyond ZK proofs—including AI computing.

This isn’t a pivot in the traditional sense. It’s a strategic expansion of the platform to capture wherever compute demand exists. They’re even using AI tools to write code that outperforms what human engineers can produce, and faster.

Building the Team That Ships

One of the most practical insights Leo shared is how he builds his team. The first members came through personal networks—friends and friends of friends with referrals from existing team members. But there’s a specific filter: everyone needs to have an authentic passion for the problem.

Cysic organizes around small groups of three to four people working on specific problems. Almost every day (except Wednesday), the team has 30-minute standup meetings where people share progress. But here’s the key: they don’t just talk about progress in the abstract. They demo the product and gather real feedback from teammates.

This creates a tight feedback loop that continuously improves both the product and the team’s ability to execute. Internal interactions become the crucible where ideas get tested and refined.

The Scalability Challenge

Building infrastructure that scales is always hard. Cysic’s approach includes open-sourcing components like the KVM (Keystone Virtual Machine), lowering the barrier to entry for developers who want to build on their network without hand-writing complex circuits.

The lesson: sometimes scaling means making your platform more accessible, not just faster.

The Takeaway

Leo Fan’s journey with Cysic illustrates something crucial for technical founders: the best ideas require constant calibration against reality. His willingness to abandon the custom chip strategy—despite significant sunk costs—and instead build a service-oriented platform shows mature judgment. He’s not married to the original plan; he’s married to solving the problem for customers.

That’s how you build infrastructure that lasts.

Want to hear more about Leo’s journey building Cysic, the technical decisions behind zero-knowledge proofs, and how he’s thinking about the future of on-chain compute? Listen to the full episode of Code Story to dive deeper into his story and the challenges of scaling crypto infrastructure.