If you work for an investment bank, and you want to see quants (i.e., your colleagues in quantitative analyst jobs) bristle with barely controlled exasperation, it turns out there are easy buttons to push: You just need to give them a complex task and ill-equip them for solving it.
So said Daniel Rosengarten, head of asset liability management (ALM) quantitative development at Barclays, speaking at this week’s conference on A.I. and data science in trading in London: “If you want to see a quant go crazy, give them a large amount of data, and Excel to work with it.”
Too few banks actually equip quants with contemporary data management tools such as Spark and Jupyter, Rosengarten added: “I’ve seen banks that have said they’ll have to upgrade their Excel spreadsheets because they’re still using Excel 2003 or 2007.” Instead of using the newest tools, there’s a tendency to assign ‘spare’ quant talent to lengthy data analysis tasks using old technologies.
This has a predictable outcome for quants. “I’ve been in quant groups where… people couldn’t work on any interesting projects, so they basically left and went back to college,” Rosengarten said.
A related quants problem is banks’ poor appreciation of the fact that, before data science solutions and machine learning can be applied to finance challenges, data needs to be cleaned. This is neither sexy nor a priority. “Cleaning the data is 80 percent to 90 percent of the work and most people don’t feel like doing that,” said Apurv Jain, a visiting researcher at Harvard University, speaking at the same conference.
Too many banks have “data swamps” instead of data lakes, agreed Rosengarten: “In the successful projects I’ve seen, the data cleaning takes years.” People need to clean the data, anonymize it, and understand its value; without this, nothing can be achieved.
If quants working on data projects in banks are to be kept happy, Rosengarten said, there are two simple rules: start small and actually complete a project from start to finish.
Not all banks are capable of this, especially in the context of quants. And not all banks are ready to play in the top ranks of data science. John Ashley, director of global financial services strategy at chip-maker NVIDIA, said only a handful of banks have made the kinds of $10 million-plus investments in hardware that really enable the application of machine learning: “If you want to hire the best data scientists you want to give them good tools – you don’t want to hire Michelangelo and give him crayons… Unless, you want a crayon drawing, of course.”
This piece originally appeared on eFinancialCareers.