Gleb Chuvpilo

On a 70-acre range near Helsinki, a beam of invisible light reaches out a kilometer and a quadcopter falls silent. No bang, no shrapnel, no missile that costs more than the thing it just destroyed. This is the demo Gleb Chuvpilo keeps giving, and it is the whole thesis of Thor Dynamics, the company where he is co-founder and CEO. The product is called Laser Armor. The pitch is one sentence long: our lasers can defeat drone swarms in seconds from a kilometer away.

What makes the sentence land is who is saying it. Chuvpilo is not a defense lifer who wandered into AI. He is an MIT computer scientist who learned to price risk on a trading desk, learned to build at Palantir, learned to back founders as an investor, and then decided the most interesting problem left was the cheap drone that can take down an expensive everything. Laser Armor is his answer: detect, track, and neutralize airborne threats with a directed-energy system smart enough to fight a swarm, not just a single target.

Modern air defense has an arithmetic problem. A drone can cost a few hundred dollars. The missile fired to stop it can cost six figures. Send a swarm of the cheap things at a power plant or an airport and the math breaks the defender, not the attacker. Thor Dynamics attacks the spreadsheet, not just the sky. A laser shot is, in effect, the price of electricity. The beam travels at the speed of light, hits with pinpoint accuracy, and leaves minimal collateral behind because there is nothing falling out of the sky except the drone it just cooked.

The hard part was never the physics of burning a hole in a hull. It was doing it autonomously, against many targets, in the chaos of a real attack, in places where the network might be jammed or gone. That is where Chuvpilo's background stops looking like a grab bag and starts looking like a plan.

Before he armed AI, he audited it. For several years Chuvpilo published the AI Research Rankings, an annual reckoning of who was actually producing the science at NeurIPS and ICML, the two most prestigious machine-learning conferences. He went paper by paper, tallied authorship shares using a fractional count borrowed from the Nature Index, and rolled the numbers up by organization. The verdict that traveled furthest: Google, through DeepMind, Brain, and Research, was the quiet superpower of modern AI. The work was widely read because it was unsentimental - a quant looking at a field that prefers to talk about itself in adjectives.

That instinct now runs through Thor Dynamics. The company joined NVIDIA Inception and built its AI stack on NVIDIA's tools. Reinforcement-learning agents are trained inside Isaac Sim and Isaac Lab, throwing thousands of virtual drone attacks at the system so it learns better engagement strategies before a real drone ever appears. On the hardware side, NVIDIA Jetson processors sit at the edge, fusing sensor feeds and making decisions in real time so the system can act even when it is cut off from the wider network. “NVIDIA Inception is a game-changer for Thor Dynamics as we continue to innovate in AI-driven drone swarm defense,” Chuvpilo said of the partnership.

Chuvpilo holds a master's from MIT's Computer Science and Artificial Intelligence Lab and an MBA in finance and strategic management from the Wharton School. He layered on a CFA charter, an Energy Risk Professional credential, and a stint in MIT Sloan's financial technology program. On paper it looks like indecision. In practice it is a toolkit for exactly the company he is now running, one that has to be equal parts laser lab, AI shop, and capital-intensive hardware business.

He started as a quantitative trader at Lehman Brothers and Goldman Sachs, where AI-driven portfolios taught him to trust models that beat intuition. He spent years investing alongside Peter Thiel at Thiel Macro, the global-macro hedge fund. He was an early engineer at Palantir, back when building software for hard, real-world problems was the whole identity of the place. Then he became a founder, repeatedly.

His startup history is a list of things that got finished. He co-founded Authy in Y Combinator's 2012 class; it was acquired by Twilio and became the backbone of two-factor authentication for a generation of apps. He co-founded Ride, funded by TPG Growth and acquired by Enterprise. He co-founded Pager, backed by NEA and Lux. Each was a different domain - security, mobility, health logistics - and each rewarded the same trait: a willingness to build the unglamorous middle of a system and ship it.

Alongside the operating work, he runs Thundermark Capital, a deep-tech venture firm that backs the kind of frontier companies he now competes to hire from. He also serves on the board of MIT delta v, the institute's educational accelerator, closing a loop back to where his own technical story began.

The AI Research Rankings deserve a second look, because they explain how Chuvpilo thinks. AI as a field tends to describe itself in superlatives - breakthroughs, revolutions, paradigm shifts. He treated it like a balance sheet. The method was deliberately boring: take the accepted papers at NeurIPS and ICML, split the credit for each paper fairly among its authors, attribute those fractions to the institutions behind them, and add it all up. No vibes, no press releases, just a Publication Index that said, in numbers, who was doing the work. The rankings were read precisely because they refused to flatter anyone.

That discipline is the opposite of how most defense pitches sound, and it is why Laser Armor is framed the way it is. Chuvpilo does not sell fear. He sells arithmetic: a swarm of cheap drones against a defender forced to spend a fortune per intercept is a losing trade, and a laser that fires for the cost of electricity flips the math back in the defender's favor. The same person who once tallied authorship shares now tallies dollars per shot, and reaches the conclusion that light wins.

There is a detail in his family history that is hard to invent. His father, Albert Chuvpilo, who lived from 1930 to 2020, was a prominent Soviet scientist who worked on solar technology for the Soviet Space Program. One generation harnessed the energy of light for a superpower on one side of a cold war; the next is aiming light to defend the other. The son does not lean on the symbolism, but it is there in the work: a family that has spent the better part of a century learning what you can do with a focused beam.

Chuvpilo describes himself as a serial entrepreneur and investor, and the through-line across the trading desk, the AI rankings, and the laser range is the same: take something the rest of the market is mispricing and build the cheaper, smarter answer. He frames it in plain terms - venture-backed companies, he has written, are a prime driver of both economic growth and private-sector employment - and in something close to a creed: we do not inherit the Earth from our ancestors, we borrow it from our children.

Thor Dynamics is headquartered in Orlando's UCF Research Park, with a European base and that 70-acre live-fire range near Helsinki, plus roots in Silicon Valley where Chuvpilo has long operated. The founding team is built for the mission: Scott Buchter, a laser-business veteran and former MIT Lincoln Lab researcher with a CREOL PhD, serves as COO; Kenneth Braithwaite, the 77th Secretary of the U.S. Navy and former Ambassador to Norway, lends the company its defense gravity. In early 2025 the company unveiled Laser Armor, raised a seed round to push the product forward, deepened its NVIDIA partnership, and took the pitch to the Slush stage in Helsinki later that year.

The bet is straightforward and large. If drone swarms are the cheap, scalable threat of this decade, then defense has to get cheap and scalable too. Chuvpilo's wager is that the answer is not a bigger missile but a smarter beam - one that learns in simulation, decides at the edge, and fires at the speed of light for the price of a wall socket. It is a strange specific of a company, and that is exactly the point.