FINALIST YEAR: 1991
HIS FINALIST PROJECT: Teaching a computer to identify cancer cells
WHAT LED TO THE PROJECT: Ciamac Moallemi always loved computers. As a kid growing up in Queens -- his graduate student parents immigrated from Iran to the US in 1978 when he was three -- he spent hours playing around with programs. His father taught mechanical engineering at Polytechnic University’s Brooklyn campus, and so in high school, Ciamac found a gig working with a professor in the school’s computer science department. They looked at a quintessential computer problem: how do you teach a machine to recognize patterns?
In this particular instance, they wanted to teach a computer to recognize bladder cancer cells in a urine sample. An experienced person can quickly spot cancerous cells amid all the background material. Teaching a computer to recognize such cells, on the other hand, involves figuring out which variables are important, and writing a program to extract the relevant information and match it up against the chosen criteria. After a good deal of effort, Moallemi taught the computer to recognize cancer cells often enough that, when he entered his results in the 1991 Westinghouse Science Talent Search, he placed fifth.
THE EFFECT ON HIS CAREER: “In a broad sense, it got me interested in this stuff,” says Moallemi – that is, selecting and optimizing variables to come up with a clean answer. “But in a more narrow sense, it got me into college” – technically as a high school drop-out. Looking at all the variables in his own life around 1991, Moallemi decided that he’d had enough of secondary education, and so he applied to and was accepted at MIT a year before he would have earned his high school diploma.
At MIT, he studied math, electrical engineering and computer science. Given his family background, he felt some pressure to go to graduate school and into academia. “I would be the only one without a PhD” at family gatherings, he explains. But he was soon lured toward finance, another field that involves recognizing variables and optimizing some of them. “You can make money doing that,” he says. He started working for a hedge fund, Delta Global Trading, during his junior year. He became a partner not long after graduation.
He enjoyed the day-to-day aspects of the job. “You get a lot of real time feedback,” he says. You come up with an idea, and within 4-6 weeks you know if it’s a good one. “It’s not a subjective assessment,” he says. “There’s a number.” You have a spreadsheet, and on it, your numbers are either black or red.
Moallemi moved onto a few different technology positions at start-ups, but eventually, he realized that for all the practical upsides, he missed theoretical work. In finance, “people care about what the answer is and how quickly you get it – there’s no premium placed on a clean conceptual model. The only people who care about that are in academia.” So he decided to become a professor which – as all his family members knew – requires a PhD. He went to Stanford and earned a doctorate in electrical engineering.
WHAT HE’S DOING NOW: These days, Moallemi is a faculty member at Columbia University’s Graduate School of Business, where he’s combining his interests in theory and practice. He studies why certain systems don’t always produce the same output for a given input. The stock market is one such system.
This leads to a variety of financial engineering problems. For instance, say you hold a very large block of stock and want to sell it. You want to get as much money as possible for it (that is, optimize your revenue from the sale). “If you sell it too quickly, you’re going to move prices,” says Moallemi. That’s because other traders will see that someone is trying to dump a large quantity of stock, and assume that there is a good reason to dump it – that it’s a bad investment. “You try to optimize how you spread it out over time to minimize how much you impact prices. But the flip side is that if you spread it out too much, other people will realize what you’re doing. So they’ll sell a little in front of you and make money at your expense.”
So what should you do? From a game theory perspective, it turns out that the answer depends on market liquidity – that is, how easy it is to sell stock. In a market where it’s easy to buy and sell, you should spread things out over time. In a more illiquid market, where it’s hard to sell, you should concentrate your trades in a very short interval of time, in order to get the best price. Yes, you will cause prices to drop, but since prices are unlikely to bounce back in an illiquid market, you will get stuck with continually lower prices if you sell your shares over time.
Given the recent breakdown in some credit markets, the latter scenario is becoming quite relevant. “Ciamac’s work is advancing our understanding of how strategic interactions among financial institutions play out as a consequence of changes in their demand for liquidity,” says Benjamin Van Roy, a professor of management science and engineering at Stanford. “Our recent liquidity crisis highlights the importance of understanding risks brought on by such interactions.”
Of course, during the recent liquidity crisis, Moallemi has come to realize that academia has another perk – beyond theorizing – over the financial world. “It’s definitely much less stressful to be an observer on the sidelines!” he says.