JOIN OUR COMMUNITY OF DATA EXPERTS

Hello everyone, it´s Louisa. I hope you are all doing well.
You are watching Meet the Expert, the live stream where experts give advice to help you successfully pursue your career.
Today I’m here with Lucile and Marion, both data scientists at Societe Generale who will talk about the challenges of increasing the number of women in data science occupations. Hi Lucile and Marion.
How are you both? Good, thanks, and you?
Good, thanks. So today, both of you will answer questions from students and tell us how to become a data scientist, the challenges and rewards of data science, and how to promote the inclusion of women in this sector.
But before we get to that, for those of you at home who would like to see more content, feel free to subscribe to our YouTube channel or follow us on Instagram by searching for Jobteaser_FR if you're looking for more tips, tutorials or advice about how to pursue your career. Before getting down to brass tacks, Lucile, Marion, I'd like to ask you to introduce yourselves. Lucile, could you start by telling us about your training and professional experience? Yes, of course. I got my master's degree in applied mathematics with a specialisation in data science. Then, as part of my degree, I did my internship at Societe Generale, after which I was hired there.
So now I have been working as a permanent employee of the General Inspection & Audit Data Lab for 2 years. Thanks, Lucile. So Marion, now can you tell us about your professional experience and how you got here?
Absolutely. I graduated from ENSAE Paris and hold a master's in mathematics of random systems (modelling).
I did several internships in trading and structuring, and then was hired by Societe Generale in 2013 as a quantitative engineer. Three years ago I got into data science and worked as a data scientist at the Investment Banking business’s Digital Office.
Thank you, Marion. We are going to take student questions now, and one of the first is, “Can you sum up what a data scientist does?” Marion, for those without any data culture, can you define the role of a data scientist in laymen’s terms? Yes, a data scientist basically is someone who retrieves data and uses maths and IT tools to make predictions. For example, Amazon and Netflix recommend books and TV shows.
Behind those recommendations is a data scientist who creates recommendation algorithms to make suggestions for you. Lucile, so that we can get a better idea and dig a bit deeper, can you give a more precise definition of what a data scientist does? Yes. Ultimately, our daily work consists of first picking a subject within the Bank that may be broad and multi-faceted, which we’ll get back to. Next comes the data. This is the real heart of the matter: how we retrieve our data in the first place, and once we’ve done that, what condition it’s in when we retrieve it. Is this proprietary data? Very rarely, so we have to go to all the effort to reprocess the data and format it, especially if it's unstructured data that we are destructuring.
Once we have that, we have our own data. Then we can apply various algorithms – as Marion mentioned a moment ago – which are usually predictive algorithms. Once we’ve done that and predicted something, we have our result: what's important is how you’ve documented and presented it. So it also consists of documentation and presentations to various audiences. Lastly, there's a chance it gets put into production.
Let's jump to another student question: “What are the skills required to go after a career as a data scientist?” Inherently, when you talk about skills, there are hard skills and soft skills. Marion, could you tell us about the technical skills required? As it stands, we have a wide variety of profiles, but in terms of technical qualities, it's true that data scientists often have a penchant for maths, coding, or IT tools dedicated to machine learning.
Overall, you have to be fairly comfortable with certain programming languages like Python, or R. And at Societe Generale, what programming language do you use? We mainly use Python for data science because Python offers a ton of open-source algorithms, meaning they are public. It offers easy access and you can reuse them, get inspired by them, and then be at the forefront of innovative algorithms by reusing those same “packages” that are publicly available.
What's more, I would definitely recommend for any student looking to get into data science to study Python. Lucile, can you talk about the soft skills, meaning the personal qualities required to be a good data scientist? One of the main qualities is curiosity. You have to be very curious about new things because data science is constantly evolving. You have to want to look up the most recent algorithms online and the latest developments. You have to be a real team player because we work a lot with other people, including data scientists and also other people who work in data. You have to have good communication to deliver these results well and be able to sell your work. And you need discipline for certain things, like using various algorithms and types of data processing, and be sure that you’ve got something accurate ultimately. Here's another question: “Which training programmes are ideal in order to become a data scientist?” Which enable you to acquire these hard and soft skills?
Marion, which training programmes do you consider to be ideal? Well, you're often looking at programmes in maths and IT generally speaking. Sometimes maths expertise outweighs IT expertise and vice versa. You do have to have some understanding of coding. Usually, the majority of data scientists have graduated from engineering schools or universities specialising in data science/maths/IT. It's true, however, that if a candidate doesn’t specialise in data science, they can have a background in maths. Nowadays you can always do online training, for example, particularly the classes available on Coursera in machine learning and AI. And then there are also platforms like Kaggle, which host data-science competitions. You can see code written by other data scientists, get inspired by it and then take part in these competitions yourself. That is a great learning experience. Lucile, anything to add?
Yeah, I'd maybe just add that you can also do a specialized master's in a year after completing 5 years of university-level studies. Marion, does the field of data science include any other occupations? Absolutely, there are many other career options in data, and data scientists even work with people in these occupations. It gives you the opportunity to learn what a data analyst does, for example. Data analysts process various types of data, such as customer or product data linked to the performance of the company. They provide indicators that make it possible to guide and run the company. Generally speaking, they are skilled with graphs and can provide highly sophisticated data visualisation tools and pretty advanced graphs. And then there are also project managers in the field of digital technology. They run projects related to the digital transformation, among others. Lucile, what's your take?
Are there other career opportunities in the field of data science? Did you want to add anything to what Marion said? Yes, to round out what she said, there is also data engineering, which would be good for those who like computers. They work more on the production side of projects and data extraction, and we actually work with them fairly regularly. Marion I know you wanted to mention another occupation? Yes, there are also UX Designers whose role is to optimise the user experience. Maybe that doesn't speak to you much, but it involves designing websites for optimal use. UX Designers provide tailored solutions for any device, including smartphones, tablets, PCs, etc., and we work with them, too.
Here's another question about your personal experience. One student wants to know, “What sparks your passion for data science and why did you choose this career path?” Lucile. Personally, there are two main things that I love about data science. The first is that it's a science that is changing literally all the time. There is always something new coming out – new algorithms and discoveries.
It's a very rich field in which you always have to keep up with what's happening. The second is the way it is applied. There are actually a multitude of applications. With data science you can work in so many different fields, and that is something really cool. Marion, why did you choose this career path? My motivation wasn't very unique – I did it for the exact same reasons Lucile did. It's true, data science is constantly evolving and that's a major draw. The projects are highly diverse. To echo what she said, you can apply it in so many different fields that it's a very interesting career. Another student is asking, you worked on recently or are working on now?” Marion, could you tell us about a project of yours? Yes, I’m currently working on a “nowcasting” project for GDP (gross domestic product) which is a very popular economic indicator.
“Nowcasting” is a cross between "now" and "forecasting". It offers immediate prediction, the goal being to predict GDP at any point during the quarter. These predictions are very useful for politicians, particularly in order to assess macro-economic conditions in real time and make crucial economic decisions. It's clear how important it is, especially during this COVID crisis.
It truly leads to decisions that affect us on a daily basis, as you can see currently. As a result, the market keeps a close watch on GDP because it needs to able to understand and see what's on the economic horizon, too. It may seem like predicting GDP is nothing new, but what is much harder to accomplish is predicting it at any given moment, factoring in a multitude of data and using innovative machine learning and deep learning tools. Lucile, can you give an example of a project you worked on recently or are working on now?
I’m going to go with a project on a subject that comes up all the time at General Inspection and Audit, which is the fight against fraud, whether you're talking concretely about money laundering, terrorist financing or even internal fraud. To shed some light on General Inspection and Audit, it's what's referred to as the 3rd line of defence. In a bank you have 3 lines of defence. The first is comprised of the teams that perform controls for their own business, their own work.
The second is permanent control, which constantly monitors what the businesses are doing. Then you have us, the 3rd and last line of defence, and we carry out periodic controls. Fraud comes up periodically, but it's fairly common, so it's a subject of great interest to us. How did we end up working on this subject? The goal is to make sure there is no more fraud in the customer transaction network.
To accomplish that, we recreated the transaction graph for all transactions between the various customers. Then we had to split this huge network into communities, which was the next step. Then, we wanted to describe each community, converting “unstructured” data into “structured” data, which goes back to what I mentioned earlier, creating one line graph per community and variables describing them. Next, we applied supervised machine learning algorithms to this structured database to predict whether a community is healthy, meaning fraud-free, or, by contrast, if fraud is suspected within it.
The goal of this project was to design a tool that could ultimately evolve and learn new fraud trends, which are constantly changing, without requiring human intervention.
When all is said and done, we still have a lot of work underway. The project is far from finished – it’s for the long term. For example, we are reworking one of the steps for creating communities by incorporating social network analysis methods. These methods are generally used for social networks such as Facebook and Twitter to split users into communities. We are also working on automating the creation of variables using graph embedding techniques. And lastly, in order to improve performance, we recently used neuron networks instead of traditional classification algorithms, which ultimately get you better results. I think one of the major challenges of the project was that our transaction graph showed very few instances of fraud. Thankfully, there aren’t that many people who commit fraud, but what makes the project so complex is how hard it is for the algorithm to learn what a thief is and what fraudulent behaviour is. Thank you, Lucile, for your explanations. Let's take another question that's addressed to both of you: "What are the current challenges for the banking sector and how do data science and machine learning provide solutions? Marion, would you like to answer this one for us?
Yes, there are several challenges currently, particularly that banks have to adapt to an increasingly strict regulatory environment while continuing to perform their key role in the economy, the central role of providing credit at any given moment. And it's even more important to be able to provide financing at the drop of a hat during a crisis like this.
Another challenge is that Societe Generale is looking to fill this role more sustainably and responsibly through ESG financing (based on Environmental, Social and Governance criteria). These 3 factors measure the sustainability and social impact of an investment in a company. How can machine learning help meet these challenges? By accurately assessing risk, providing tangible data for assessing these risks, identifying the companies and people experiencing hardship, especially during a crisis, and catching these risks early, and for everything environmental and social, by quantifying environmental and social criteria within companies using data. Lucile, would you like to weigh in?
Yes, as a matter of fact, General Inspection and Audit carried out a CSR assignment and we worked on a data science project involving the reputations of these companies in terms of CSR and climate risk.
Marion did you want to add something?
It's true that companies are increasingly geared towards innovation, which is also one of Societe Generale's core values. It's clear that this is another area in which machine learning can transform businesses, automate processes and provide increasingly personalized services to customers. I would say that the good news in the financial sector is that we possess vast amounts of very diverse data, because that's part of the culture to collect and analyse data. That inherently enables us to meet all of these challenges.
As I mentioned at the start, you were going to tell us about efforts to include more women in data science careers.
One question we’ve gotten several times is: “Why is it important to talk about bringing more women into data science careers?” Marion?
It's important because diversity, or rather the lack thereof, can hurt companies. A recent study from the BCG shows that only 15% of data scientists today are women, which is still far too few. This lack of diversity, whether it be gender diversity or other forms of diversity, can create biases because the way we think is clearly dependent on our experiences. So if we all think the same way, we will process data the same way, whether consciously or subconsciously, without really looking to challenge the way we process it. Somewhere along the way, this can lead to societal gender biases in artificial intelligence. That is why it is critical to have diverse teams in data science as well as other fields. This lack of diversity can also hinder innovation.
Simply put, the more diverse the team, the more innovative solutions we’ll have. People will think differently, so we will inherently promote innovation. There are also various studies that demonstrate that teams are generally happier and more productive when they are more diverse. I think there was even one study by the International Labour Organization that showed that sales performance improved when there was a diversity policy in place. So why hold ourselves back? All of the evidence is in favour of it.
We emphasise that, assuming equal skill levels, obviously, we would go with the more diverse choice.
But it's important to keep that in mind: assuming equal skill levels. We have to keep promoting the inclusion of women. We had another question on this subject. Lucile maybe you can answer this one for us: “In your opinion, what is the underlying reason there are so few women in the field of data science?”
Well, perhaps one reason is that, according to the BCG study Marion mentioned earlier, only 35% of female students study science, technology, engineering and maths. So there's one bias that means there are too few female students in the first place. Secondly, I think that we at Societe Generale have an additional bias due to the fact that we deal in finance. When you mix data science with finance, you have even fewer women. I see it in a class I teach at Université Paris Dauphine for their master's in finance, but the class is on Python, so it's more about data science, and there is only one female amongst my directed studies students.
I think that says a lot. Marion, would you like to add anything? I think that this is also one of the reasons why we are here today.
It's important to demonstrate that this is not a job for men only, and that any women interested in this field shouldn't think, “Yeah, okay, but working in data science at Societe Generale means growing in an environment that doesn't promote the inclusion of women at all.” It's very normal to have this kind of stereotype in mind – we get it – but it's not true. The fact that we are here today is a testament to that fact.
Well, Lucile, you were saying that there's only one female student in your class at Paris Dauphine. On your team at work, are you the only woman or are there others? No, not at all, on my team there are almost as many women as men. That's also what I wanted to impress upon the female students listening: if you come to work for Societe Generale, you will not be the only woman there, nor will you be the only man there for our male students who are listening. Let's take another question on this same subject: “How can women be made to feel that they truly belong in this sector?” Marion? Our legitimacy is not in question. We have to start by reminding women that of course they belong, no doubt about it, and that we are thoroughly convinced. Lucile?
To add to that, it goes back to what I was saying about equal skills.
We got these jobs based on our skills, not because we are women, and it will be the same for any female students listening as well as for male students. For the women out there, believe in yourselves and know that it leads to a wealth of perspectives as Marion pointed out previously. It's true. For the last question, one listener would like to know: “What can we do to include more women in this field?” Marion, are there organised efforts happening?
Yes. It involves communication – communicating the fact that we believe in diversity, and that we want them to join us. This is what we were saying before: our presence is also clearly one of the objectives here. And yes, we are also lucky to pursue careers at a group where the subject has gathered momentum, and it's an area that has clearly been put in the limelight. We have the ability to speak up and brainstorm about these issues. Recently, for example, I was able to participate in the "AI Daring Circle” collaborative working groups by the "Women's Forum”. It may not be obvious what it is by its name, but essentially these are cross-sector working groups of representatives from the realms of business, politics, NGOs, science, etc.
The goal is to get involved and make a long-term, positive impact in areas where women are disproportionately represented and where it is crucial for women to lead. So for us, it is clearly important to make our contribution to the cause, and of course it's gratifying on a personal level to be able to collaborate in these kinds of working groups. Apart from communicating about these things, it's also about promoting our goal and showing that we are here, front and centre with our expertise, and that one of the reasons this expertise is recognised is because our teams have a culture of inclusion. Lucile, are there other initiatives happening?
Yes. We speak at several multi-disciplinary events to show examples of women at Societe Generale. Today we are participating in the JobTeaser broadcast with you, Louisa, and in September 2020 we did a meetup on “AI for Finance” with the "Women in Machine Learning and Data Science” community. We spoke alongside several other female employees, one of whom is doing a research-based dissertation on natural language processing.
If you are interested in that meetup, and by you I mean students, the video is available on YouTube. We also participate in networking events and student forums where we show that there are also women that are on the teams. And Marion, is it important to promote this occupation to women fromthe earliest age possible? Yes, of course, it's very important to children and teens to explain who is behind Netflix and self-driving cars.
The idea is for them to have a professional goal at the youngest age possible without waiting until later or maybe until it's too late to ask those major questions about the direction they want to take. Any last comments from both of you for young female students looking to become data scientists? Any last few words of encouragement or anything you'd like to say to them? I'd like to tell them to believe it's possible and go for it. You can do it. Marion, I imagine you share this sentiment? Yes, absolutely. Well, that wraps up this edition of Meet the Expert.
Thank you so much Marion and Lucile for sharing your day-to-day, your challenges and especially for giving the students all these tips. For those listening at home, I mentioned this in the beginning but if you are interested in more content, tips, and tutorials for your professional career, feel free to subscribe to our YouTube channel or follow us on Instagram by searching for Jobteaser_FR.
Marion, Lucile, thank you so much again for being with me here today.
For those of you at home, see you soon for another live broadcast!