The banking jobs: Transformation project consultant
Amid rapid changes in markets, sectors and new technologies, the development of BNP Paribas'...
Machine learning, deep learning, artificial intelligence—Julien Dinh, Senior Research Lead at Global Markets, juggles new technologies like these on a daily basis. His role is to develop algorithms to optimize the performance of the activity’s sales teams. Let’s meet this tech-savvy guru.
In my role, I co-lead the Data & Artificial Intelligence Lab at Global Markets, the capital markets activity of BNP Paribas CIB (Corporate and Institutional Banking). We research ways to use machine learning technology to improve customer relations. For my part, I manage a team of five specialists in charge of leading quantitative research projects in this field. Our role is to help sales representatives better respond to customer needs by predicting behaviors and automating certain interactions.
We analyze raw customer text data—requests sent by chat or email—using a technology called natural language processing. In practical terms, we develop algorithms that can understand and interpret these written requests, which are usually about buying or selling assets. This enables us to expand our knowledge of customer needs and to automate the processing of certain requests, with no intervention by an advisor.
At the start of my career at BNP Paribas , in 2003, I was doing quantitative research on models for evaluating interest rate products. In 2015, together with my manager, the head of quantitative research (Joe Bonnaud, Global Markets), we came up with the idea of creating an applied data analysis lab for Global Markets. The activity manages thousands of daily transactions focusing on a wide range of asset classes . The intrapreneurial adventure that began in Frankfurt has now expanded to six cities, to cover all of our markets—London, Paris, Singapore, Frankfurt, New York and Mumbai.
Banks have always invested in information systems, especially for their activities involving massive data flows, such as trading and trading floor activities. These systems generally process most data effectively, except in areas that are difficult or even impossible to code, as is the case with natural language. That is where machine learning comes into play. This technology allows computers to learn on their own. In the context of our activities at the Lab, the machine gradually learns to understand the content of the written message. This process contributes to our specific goal of assisting sales teams with their daily tasks and improving service quality for our customers.
Deep learning involves using dense artificial neural networks, capable of integrating a vast number of parameters, but our work always begins with machine learning. Once we reach the limits of machine learning, we turn to deep learning expand our work.
You have to know how to code concepts and have a mastery of automated learning methods. Most of our specialists have also studied mathematics. In terms of personal qualities, determination is essential: 90% of the job consists of creating something that did not previously exist! If people can perform a task, then we need to find a way for the program to do it, too. People are always our starting point.
In Frankfurt, where I am based, some of the teams wanted to adopt the technology just to try and prove to me that it wouldn’t work. Much to their dismay, it worked marvelously—our behavior predictions were even confirmed by our customers themselves! Ever since then, our specialists have a nickname for these machines—Minions, like the little yellow creatures in overalls from the animated movie.
Photos : header ©jotily // ©metamorworks