Armando Gonzalez has got a problem. The CEO of Ravenpack, a company which uses artificial intelligence to turn news and social media into usable indicators for financial services firms, is struggling to hire one particular species of coder: Lisp (formerly known as LISP) specialists. There just aren’t enough of them to go around.
“We’re very actively hiring in data science,” says Gonzalez. “Our team is going to grow by 20% to 30% this year and we’re looking for data science specialists and developers who work across everything from Python to Lisp, but Lisp professionals are very hard to find.”
Lisp is nothing new. One of the oldest high level programming languages, it was originally specified in 1958. This being the case, you might think the world would be awash in Lisp talent. Unfortunately not: far more people today are familiar with contemporary programming languages like Python, and Gonzalez says the comparatively few Lisp specialists in existence are hired by the National Security Agency or other areas of the intelligence community which need hardcore natural language processing skills (NLP).
Lisp was the traditional language of choice for fast-growing NLP functions. Although Python and Java have mostly displaced it, Lisp remains popular for its ability to integrate macros that allow control over when and whether arguments are evaluated. Lisp is therefore still one of the best programming languages for NLP-focused AIs and is often still used for functions that need to parse large quantities of information.
Parsing large quantities of information is Ravenpack’s specialism. Having started out in 2007, the company is one of the longest established players in the market for “machine readable services.”
“We’ve solved many of the problems inherent in processing natural language,” says Gonzalez. “Common Lisp remains one of the top languages for artificial intelligence applications… it’s a powerful text processing tool that allows our team to quickly experiment and build analytical prototypes. Today’s Lisp compilers are robust and flexible and allow development entirely within Lisp in combination with other languages like C.”
Natural language processing is problematic in itself, but in financial markets Gonzalez says NLP needs to be overlaid with a framework that interprets the linguistic information. “We’ve spent the last 10 years building a database that allows the computer to read and to understand, for example, that people are talking about Apple the company rather than apple the fruit,” he says. “We’ve also built a series of programs that can understand when people are talking about particular topics, and whether they’re doing so bullishly or bearishly. The programs interpret that information based upon how the market has reacted or is likely to react. We have 17 years worth of millisecond data to consult – that’s a very detailed backtesting set and it gives our clients in banks, hedge funds, and asset management firms an edge.”
Gonzalez offers his programmers a fairly rarefied life. Ravenpack has offices in New York and London, but its data professionals and developers are based on the South coast of Spain. “We try and offer people something different by asking them to join us in our Marbella R&D centre,” says Gonzalez. “Everyone we’ve hired in the last 10 years has found it to be a very good quality of life here – much better than London or New York. Your salary buys a lot more and there’s a large expat community with some excellent schools. You have the beach nearby and snow a few miles away on the Sierra Nevada.”
Gonzalez says it’s easier to train developers in finance than it is to train finance professionals in software development. “When we’re hiring developers, they don’t necessarily need to know about finance: it’s all about coding ability, speed of implementation and design.” In data science, Gonzalez says familiarity with econometrics is a good thing, but that banks and hedge funds are competing for the same talent: “We are happy to educate excellent data scientists in finance.”
Should non-technology professionals be apprehensive about Ravenpack’s expansion? Gonzalez says researchers who rely upon manual analysis probably have most to lose. “The market is changing. There’s less and less demand for fundamental trading strategies and more demand for quant investment funds. It’s the quant funds who use our product, and they’re displacing the old manual analysts.”