Using Natural Language Processing, Qunatxt Aims To Put Computer Science And Analog Finance On The Same Page

With the second annual Benzinga Fintech Summit in the books and another year of events brining together luminaries from across all aspects of finance, we're taking the opportunity to debrief with a few of the participants to get their opinion on the current state of their industry and to get a survey of the changes they expect to see in the near future.

In this interview, we spoke with Matin Kamali, CEO of data processing and extraction firm Quantxt.

Tell us a little about Quantxt's history in the fintech industry, particularly its outlook on the processing and efficient delivery of quant data.

The first but a very essential step in building models is feature extraction. It is about identifying important factors in data and extracting them into a format that can be fed into a model. There is no textbook that formulates the process of feature extraction from a raw data source. You can only discover them experimentally. In the case of working with large data or unstructured sources, such experiments can be very time-consuming. You end up building a number of custom feature extraction modules which, most likely, won’t be re-used in any future experiment.

We provide a solution for discovery and extracting features from data sources in an efficient way so analysts and researchers can spend more time on improving the core of their models.

What trends in fintech and data processing is Quantxt paying the most attention to right now?

On average 2% of a mid-size or large financial [institution] consists of Traders or Portfolio Managers, while Analysts or junior Associates account for about 70% of the workforce in such firms. (Study was internally done on LinkedIn public data)

The 2% play the most critical role in a firm. Every single decision they make on a daily basis matter the business. The 70% is the hard-working soldiers behind the scenes. They prepare summary reports, do a number of manual and, sometimes, repetitive tasks on daily basis. They maintain and run financial models and produce spreadsheets of analytics to be reviewed by the decision makers. They usually don’t make any business decision but their function is vital for the firm.

I have been part of the 70% in one of my past careers, and I am very interested in bringing efficiency to them. Our technology is aimed to automate a number of data processing tasks that are done manually by financial analysts.

What do you see as the major challenges and opportunities arising in the creation and delivery of financial data over the next five years?

Investment professionals are becoming more interested in using alternative data sources, beyond traditional fundamental and market data, to find investment opportunities. As part of that movement, a large number of alternative-data vendors and alternative-data marketplaces have been formed in recent years. One major challenge is that the providers, who are often data scientists and engineers, and the financial professionals don’t understand each other’s world. For example, the concept of Benchmark and Testing in the world of data science is so different than how it is defined in the world of investment.

There are a number of on-going efforts to standardize the way data vendors offer their products. Through it either requires the buy-side to provide some guidelines on how they would want the data and/or the vendors to all to share a common practice on how they want to deliver their data. This is not an easy problem at all. Sharing expectations from data by buy-side could reveal their trade secrets and agree on a common practice on data delivery may put a burden on vendors around innovations in building data-sets. It will be very interesting to see how this domain evolves in the next few years.

What advice would you give to someone with an idea that could disrupt financial services?

I had always thought the financial industry uses the latest technology. In reality, it was quite the contrary. Spreadsheets were (and still are?) a major part of almost every decision-making process. This was an interesting observation right after I started my career in finance. Before that, I had the opportunity to work with some of the top brains in the area of computer science in an environment where decisions were made based on the results of models run on thousands of parallel CPUs.

Spreadsheets won’t go anywhere anytime soon. They are simple! You don’t need to attach a “How-to” sheet that goes with it, which is so unlike a number of AI-enabled tools.

My biggest advice is to try to understand transparency and simplicity of financial workflows before building new ones. You may think twice on how you want to explain your process to your potential users. Of course, you need to solve a business problem but transparency is what you need to prove before convincing your users to try your product.

Check out the Benzinga Fintech Summit website to see who else made this year's gathering the best yet and to view videos and pictures of the event.