From Harvard at 18 to Building Crypto's Fastest Exchange | Vlad, Lighter

Vlad qualified for both the International Physics Olympiad and the International Olympiad in Informatics teams, making him one of the first people in the US to achieve that combination. His training camp cohort in the early 2000s included two future Anthropic co-founders, the person who became Facebook's first CTO, the current US vice president, and founders of several large quantitative trading shops. He attended Thomas Jefferson High School in northern Virginia, where he also met a future Robinhood co-founder.

He finished high school at 16 and completed Harvard in roughly two and a half years by converting approximately 12 AP credits into a full year of standing and taking six courses per semester. His reasoning was explicitly economic: undergraduates pay tuition while graduate students are paid, making the fastest transition to graduate status the strictly dominant choice. His pivot toward applied work came during the experimental component of the Physics Olympiad, where he ranked second in the world despite a poor theory performance, and that hands-on encounter with real data oriented him toward finance and trading rather than academia.

At Harvard he traded from his dorm room attempting to emulate Ken Griffin and Jim Simons, finding strategies with strong back-tests but near-zero live results. Griffin later told him directly those strategies had already been arbitraged away and argued that finding alpha in nearly efficient markets requires combining multiple perspectives and is better done in a team. Vlad spent roughly one year at Citadel before leaving to build a trading desk at Graham Capital from scratch, which he describes as his first experience leading a team inside a larger institution.

He moved to Silicon Valley in 2012, spending about a year and a half at Quora on machine learning before becoming head of all engineering at Addepar, a firm founded by Palantir co-founder Joe Lonsdale on the thesis that the miscalculated risk of 2008 resulted from unconsolidated financial data. Peter Thiel, whom he met through the person who became Facebook's first CTO, advised him not to start a company immediately but to first learn how startups operate. His exposure to what is now called AI predates Silicon Valley by roughly two decades: his Intel Science Talent Search project used optics to run neural networks faster, a concept now called photonics, and Nobel laureates he met through that competition told him neural network research was a dead end and advised focusing on quantum computing instead.

He read the Bitcoin whitepaper in 2012 and had crypto on his shortlist in 2017 when he chose to found Lunch Club instead, concluding the technology was not ready, in the same way he believed 2014 AI was not ready for the applications he had in mind. Lunch Club, co-founded with Scott Woo who is now co-founder and CEO of Cognition, was a professional networking product that used AI for matching and avoided adverse selection by having the system make matches rather than relying on double opt-in. The market shrank after COVID, and during the 2022 pivot the team ran an internal YC-style process testing three ideas simultaneously. Lighter won that process, retained 80 percent of the engineering team, and operated as the same legal entity as Lunch Club rather than a new company.

The core problem Lighter set out to solve is that most digital assets trade on centralized rails despite representing decentralized protocols, with decentralized AMM volume representing less than one percent of total digital asset trading when Lighter began building. Vlad identified five pillars no existing exchange satisfied simultaneously: low cost, low latency, secure, verifiable, and composable. Centralized exchanges like Coinbase and Binance cannot offer the auditability that ZK circuits provide, and Uniswap cannot offer comparable speed and latency. The technical solution required 18 months because it had never been done before, and Vlad describes ZK proofs as a once-in-a-generation improvement similar to the transformer in AI, improving performance across multiple dimensions relative to the prior frontier simultaneously. Lighter targets perpetuals specifically because decentralized perps that are secure and verifiable is something traders want but no one had previously delivered, and Vlad argues verifiability would have prevented scandals like FTX and Madoff.

Lighter's long-term mission is to serve as the technology layer for the merger of DeFi and TradFi, with a composability frontier covering tokenized stocks, perpetuals, and options sharing the same balance sheet. On TradFi adoption, Vlad is explicit that large institutions will not switch infrastructure based on claimed improvements alone but will test new technology in pockets before broader adoption, constrained by shareholder obligations and management structures. Lighter's strategy is to engage TradFi as willing partners and allow adoption to grow from those margins inward, with current partners including Telegram Wallet and Robinhood. Founders Fund co-led Lighter's last venture round before its token launch.