Quants, with their computer, math and statistical modeling skills, are as sorely needed as ever in financial services, but their role has changed dramatically since the heady precrisis days.

Instead of pricing complex derivatives and helping to fuel exotic markets, they're more likely now to apply their massive brainpower to comparatively mundane activities like risk management and stress testing.

At Carnegie Mellon University, only 45% of the December 2014 graduates of the school's quant program (it's technically called Master of Science in Computational Finance) moved on to front-office trading roles, down from 60% a decade earlier, said Steve Shreve, the school's Orin Hoch Professor of Mathematical Sciences. By contrast, a quarter of last year's class went into risk management, up from 15%.

"These days the big banks have consolidated and quant jobs there are much more structured," said Emanuel Derman, head of risk at Prisma Capital Partners, a professor at Columbia University, one of the original quants and author of "My Life As a Quant: Reflections on Physics and Finance." "There's lots of regulatory work, lots of 'model risk.' It's not as exciting as in the old days, I believe."

The changes reflect the industry's postcrisis emphasis on playing defense, also borne out by who else banks have been hiring (compliance specialists, largely) and where they spend their tech dollars.

Where the Jobs Are

If you do a Google search for quantitative analyst jobs at banks, most of the results are for jobs in risk management.

"Risk has really come on the scene since the financial crisis. It's a growth area," said John Lehoczky, professor of statistics at Carnegie Mellon. "Risk managers call the shots in a way they didn't before 2008."

In the past, "every time the banks would have a crisis, the regulators would say, 'why don't you get more serious about risk management?' and it would last for about a year. This time it's lasted longer."

One of the skills quants develop to evaluate risk is the ability to do Monte Carlo simulations, a modeling technique used to approximate the probability of certain outcomes by running multiple trial runs, or simulations, using random variables. To calculate loan portfolio risk, they tend to use a continuous time model developed by Robert Merton, according to Shreve, who teaches a course on portfolio risk modeling.

There's also been a rise in demand for quants in algorithmic trading, Lehoczky said. Trading is generally considered the most exciting area for young quants to get into. "And there are still bonuses paid," added Shreve.

Indeed, one constant is that the quants are still very employable in the financial sector.

"The demand has far exceeded the supply," Shreve said.

Stress testing is another area where banks have been hiring more quants, to help them predict their ability to withstand the government's worst-case scenarios.

Regulators have also been seeking to hire more quants, but have a harder time than banks do. Naturally, the government has a lower pay scale than large banks.

"The government comes knocking," said Richard Bryant, executive director of the computational finance program at Carnegie Mellon. "Our students have good offers from non-government [employers], and for the same reason they often prefer the front-office roles, they prefer Morgan Stanley over the Fed. But the Fed does try. They're hiring."

Computer Skills Critical

Programming skills have become more important for quants.

"When I first came to Wall Street, it was important to be a good programmer and do your own 'dirty' work," Derman said. "Then that became unnecessary. Now, with algorithmic trading, that's important again."

Lehoczky also sees a greater emphasis on computer science and statistics. "The quants today need to be self-sufficient, testing their own strategies and able to gather data," he said. "That's a whole skill set that you just can't hand off to someone else, give to the IT department. There's no question that the computer scientist is much more highly valued today than has ever been the case."

The statistical machine learning portion of Carnegie Mellon's program for data scientists has become more popular.

"Data scientist is probably still hanging in there as the sexiest job, it's a very lucrative field," Lehoczky said. "We see a rise in that skill set. The modern statistician is one who has incredible computing skills and is able to handle massive data sets."

Carnegie Mellon also has a course that teaches quants to stand up and sell themselves, to make a presentation. And a lot of time is spent on ethics. Finance "is one area where it's important to be aware of ethics," Lehoczky said. "The stories you read tend to be [about] unethical [conduct]."

Don't Look at Us

Right after the financial crisis, some people wondered why the quants and their predictive models had failed to anticipate the meltdown.

According to Shreve, the models were being used in places where their underlying assumptions were not being met. The banking industry got ahead of its models and should have acknowledged it had entered new territory, he argued.

"You should think hard about what size position you want to take," he said. "Do you really want to lever this position up based on a model that's being stretched beyond its limits?"

When traders are separated from the model builders, there's a problem, Shreve said. "The traders say 'yes, this is working, let's make money,'" he said. "And the model builders are trying to put on the brakes and don't have the clout."

And when the quants were pricing new collateralized debt obligations and securitized mortgages, the model they used was the Gaussian copula, which is good only for pricing, not for hedging, Shreve said. "So you can compute a price, but if you don't have a hedge, you're exposed, and somehow this was not taken into account because there was too much money to be made," he said.