Bad Behavior Database Aims to Stop Rogue Traders Before They Actby and
Goatherd who made it at Goldman uses AI as surveillance tool
Behavox challenge is getting banks to share information
You’re young and want to get ahead. First in and last out of your Wall Street office. You work weekends and message colleagues with inspired trading ideas at all hours. Guess what? You’ve just marked yourself as a potential rogue trader.
Welcome to the brave new world of trader surveillance, where former Goldman Sachs Group Inc. research analyst Erkin Adylov is building a library of banking villainy based on the behaviors of hundreds of past miscreants such as UBS Group AG’s Tom Hayes and Societe Generale SA’s Jerome Kerviel. Using thousands of inputs, from stress levels in voice recordings to the frequency of visits to the staff canteen, Adylov and his team at startup Behavox grade employees on how likely they are to go bad before they do anything wrong.
While that may sound like something out of a Philip K. Dick science-fiction novel, hedge fund Marshall Wace and interdealer broker TP ICAP Plc are already using the software to monitor employees, and some of the biggest investment banks and commodities dealers in the world have begun testing it. After paying fines of more than $200 billion in the past eight years for a catalog of failings from money laundering to market manipulation and terrorist financing, banks are looking to companies like Behavox to keep them out of regulators’ crosshairs.
“If you don’t know what your employees are doing, then you’re vulnerable,” Adylov, a 33-year-old rail-thin native of Kyrgyzstan, says in the rapid staccato typical of traders for whom a wasted second is a lost opportunity. “Some banks don’t seem to want to know how exposed they are, and they are the ones who are going to get fined next.”
Under scrutiny like never before, finance firms are spending as much as one-fifth of their revenue on compliance, hiring tens of thousands of investigators, out-of-work traders and former intelligence officers to sift through employee communications. It’s not just the threat of fines that’s motivating them: Under new rules that came into effect in March, senior managers in the U.K. can be held accountable for the actions of their underlings and even face jail sentences.
Behavox uses machine learning, also known as artificial intelligence, to scrutinize every aspect of an employee’s working life. The technology enables computers to teach themselves how to collate and analyze huge volumes of data. Behavox scans petabytes of data, flagging anything that deviates from the norm for further investigation. That could be something as seemingly innocuous as shouting on a phone call, accessing a work computer in the middle of the night, or visiting the restroom more than colleagues. The system checks these behaviors against case studies of past traders who have strayed from the straight and narrow and looks for a match.
While other companies use similar technology to watchdog trading floors, Behavox takes it one step further, compiling a central repository of behavioral patterns accessible to all clients. Adylov calls it the Conduct Risk Exchange. His challenge is to persuade firms to share potentially embarrassing details about their inner workings. If he can build a network big enough -- more than the three companies that have signed up so far -- it could change the way banks police themselves.
“With this system, users benefit from eyes and ears across the industry and the behaviors that are being spotted and fed into it,” says Conor Kiernan, chief technology officer at Marshall Wace, a $28.5 billion alternative asset-management company based in London. “It really doesn’t pay to be insular when it comes to compliance.”
There are signs larger lenders are loosening up. Barclays Plc, HSBC Holdings Plc and other major European banks last year formed the Cyber Defence Alliance to pool information about hacking attacks. Still, it may not be easy for an untested entrepreneur to sell Goldman Sachs on the idea of trusting him with something as critical as compliance.
That’s not the only hurdle Adylov faces. The regulatory crackdown that has reshaped finance since the crash may ease with the election of Donald Trump as U.S. president. The billionaire has pledged to roll back the 2010 Dodd-Frank Act. If scrutiny of banks relaxes that could curtail spending on compliance.
“Six years of legislation would take time to unwind, but the tone at the top in the U.S. has changed dramatically,” says John Harvie, a director at consulting firm Protiviti Inc. “This could change the way regulators are able to operate, and not just in the U.S. Under Brexit, the U.K. could also end up with less regulatory pressure as London tries to compete with other markets.”
None of that daunts Adylov, who’s no stranger to adversity. Born to farmers in a small town in what was then part of the Soviet Union, he was a goatherd until he won a George Soros grant at the age of 16 to study in the U.S. for a year. Later his parents, who earned $120 a month, sold their home for $2,300 to pay for a one-way plane ticket to Hawaii for their son, who had secured a scholarship at Hawai’i Pacific University.
Adylov worked four jobs to fund his studies, including one selling U.S. flags and another as a gardener for money manager Clinton Bidwell, who paid him $100 a month with free room and board. Impressed with Adylov’s intelligence and charm, Bidwell took him on as a research assistant. The complexity and technical nature of the job appealed to his analytical mind, and soon the only thing he wanted to talk about was company stocks.
“It was almost like I discovered a new world that I didn’t know existed,” Adylov says from his office in the scruffy Vauxhall district of London, across the train tracks from the imposing headquarters of British intelligence agency MI6.
Despite working 60 hours a week and sleeping as little as three hours a night to squeeze in some studying, Adylov graduated with a 4.0 grade-point average. That landed him a scholarship at the London School of Economics, after which he joined Goldman Sachs in 2007 in London as an equity research analyst. Fast-forward seven years and Adylov was a top portfolio manager at hedge fund GLG Partners earning $1 million a year.
Married and with a child on the way, his future looked secure. That’s when he quit.
“I realized that if I scrolled my life forward 20-30 years it was highly predictable,” Adylov says. “My life has always been a chase, a perennial progression toward something better. For me, it’s all about the struggle.”
Adylov set up Behavox in June 2014 from his apartment in south London. His vision was simple: Bring open-source collaboration and machine learning to the world of bank compliance.
Over Skype, he recruited his chief technology officer and co-founder Roman Zelov from Oracle Corp., where he was principal software engineer in St. Petersburg, Russia. Adylov moved Zelov into his cramped apartment, where he spent five weeks sleeping on his sofa.
Adylov only hires people from similar backgrounds, so the two had a lot in common. Both are overachievers with humble origins. They hit it off, and that December started to assemble a team. It took another 15 months before they had a finished product.
The software works by analyzing employee behavior based on historical observations of individuals and their peer group. It detects anomalous conduct and gives it a risk score, derived from that person’s past compliance record. The system also maps relationships between employees and flags potential problems.
Behavox has built algorithms based on 16 years of publicly available enforcement cases against traders and banks all over the world. Clients use these scenarios to analyze behavior in their own shops. When one of them discovers an employee has done something wrong, Behavox encourages the firm to share details on an anonymous basis.
“The more firms that get involved the faster the community will be able to adapt to changes in the marketplace,” says Marshall Wace’s Kiernan. “Using tools like Behavox we can work as a community to identify behavior and share insights.”
Hussein Kanji, a partner at London-based Hoxton Ventures, met Adylov in late 2015 and heard his pitch. Kanji was confident Behavox could become a fast-growing vendor of surveillance tools, but he was more intrigued by the idea that the Conduct Risk Exchange could become an industry standard. Something else about Adylov impressed him.
“He never wants to spend money,” says Kanji, whose fund led a debut investment round in Behavox in July.
When Behavox held an open house in October, Adylov and his team did the catering themselves, serving up homemade lamb chops and chicken salad. There were no foosball tables or other staples of a startup workplace. As clients mingled with members of the Behavox team, Adylov pointed out a pair of developers hunched over their computers. “They have a deadline,” he said, half-jokingly.
Behavox promises to save money for financial firms that have spent hundreds of millions of dollars hiring lawyers, accountants and back-office personnel to comply with rules around risk, capitalization and money laundering. It and other new tech firms can automate many of the same functions for a fraction of the cost.
“Even if regulations are relaxed, the business case for reducing the expenses of compliance will remain very compelling,” says Mike Baxter, a partner in Bain & Co.’s financial-services practice in New York.
But Adylov’s ambitions go beyond saving banks money. He aims to do nothing less than change how the global financial industry ensures it’s playing by the rules.
“These firms are all over the place, all doing compliance differently, with no way to collaborate,” Adylov says. “These guys just can’t solve the problem on their own.”