Big banks have been beefing up their compliance efforts in recent years, hoping to both appease regulators and avoid another financial crisis (not to mention damaging lawsuits due to unsavory trading activity). But most financial institutions' attempts to catch wrongdoing are reactionary--by the time analysts or algorithms have flagged something as suspicious, the deed has already been done. Digital Reasoning's goal is to stop rule-breaking before it happens. Its cognitive computing software, Synthesys, does this by emulating human learning processes and applying them to the data it evaluates, putting it into context. A suspicious email won't just be tagged--Synthesys will also take into account previous correspondence, as well as current events that may be relevant. As a result, the technology is able to detect someone's intentions and prevent crimes before they happen. Digital Reasoning has been working with the federal government and applying its software to military and counterterrorism intelligence efforts for more than a decade, but in the past few years, it realized it could also be beneficial to Wall Street behemoths. As its software continues to improve--it trained the largest neural network to date in July 2015, with 160 billion parameters, beating Google's record of 11.2 billion--the company has plans to apply its technology to other areas, including financial markets, health care, and energy.