In our long-running series, "How I'm Making It," we talk to people making a living in the fashion industry about how they broke in and found success.
When you visit Lyst and accidentally misspell a designer name or the word "mules," the website will pull up a page autocorrecting the error, or if you start typing "high waisted jeans" in the search bar, Lyst.com will predictably fill out the rest of your query before you do, like a fashion-conscious Google. All of those minute, convenient details are thanks in part to data scientist Sandra Greiss, who's been with the company's engineering department since 2014.
With Greiss's education — a bachelor's degree in physics from Paris, and a master's and Ph.D. in astronomy and astrophysics from the UK — a path towards academia or a job in finance seemed like the safest bet career-wise. "I never thought I would end up in the fashion industry," says Greiss. "Most physicists don't and to be honest, it's very frustrating. It's great that we're trying to make a point that there's so much you can do with your skills."
We spoke with Greiss about her role at Lyst, why data science is important for fashion and how you, too, can get a job in the industry doing a not-so-traditional fashion gig.
What exactly do you do as a data scientist at Lyst?
Basically we use data — lots of it — to build algorithms and models to train a computer to do a lot of basic things that you can get as a human. In the case of this, with millions of products on Lyst, you need that. We can classify or predict something, depending on your problem. Say, in our case, we've got images of products and we basically build a model and train a computer to be able to detect things about products. Recently, we managed to figure out styles, whether it's a bag or shoes. The latest thing I've been trying to work on is to get a model to detect material from an image, whether it's a leather jacket or pair of denim jeans.
What does Lyst learn from data science?
Lyst is an e-commerce platform, so it's aggregating millions of products from high-end to high-street brands. We mostly use data science to make sure we recommend the most relevant products to our users, depending on what they click on and where they come from on the web. Since there's a vast amount of products in our database, we want to tailor our user's experience according to every single one of their needs, which could be through personalizing the website for each user.
What was the interview process like to get the job that you have now?
You do a coding test to show your coding skills and if you understand the basic stuff. But the type of test depends on the level you apply for, so the more senior you are, the more complicated the questions will be. The coding test will have you write a piece of code and explain what you tend to do after that, which allows recruiters to see how you approach problems and typical issues as a data scientist. My own interview was a bit informal, so the test was done on a white board — now, it's on a computer. I actually didn't manage to figure it out but they liked the way I was approaching the problem. They could still see what kind of person I was.
What projects are you most proud of?
For me personally, the more exciting projects I worked on was autocomplete and autosuggest. We call a search for something on the website and it's trying to guess what our users are after. That's a project that I was given last summer to do on my own from beginning to end, which is exciting. I thought it was a good challenge in a sense that I get to learn a lot of things and then see it go live.
If you made a spelling mistake in your search, we wouldn't be able to return results unless we had the right search query correctly spelled. So we created a service called 'Did you mean this?' to correct spelling mistakes. That's something else that I also built.
How long do these projects usually take?
They take about two to three months from beginning to end to complete because gathering the data takes time, and then you're cleaning it up and building the model and finding different ways in approaching the problem as well. Speed makes a huge difference; finding something that will perform better and quicker.
What advice would you give to someone who's interested in a similar career path?
The thing is, it’s not impossible. We have engineers that basically went from 12-week courses to jobs straight away from that. If you're interested in it, you have meetups and online courses. It's great that there are so many open courses now to teach you code and understand the concepts behind it.
Why is data science important for the fashion industry at large?
A lot of big fashion brands have e-commerce platforms, which generate a lot of data for data scientists to exploit. We use data science to understand the customer's needs and study their behavior. This is crucial for the fashion industry because brands are able to increase sales with better services that predict needs and help their customers find what they are looking for. Data science is also used to forecast the lifetime of a product on the website and advise customers how likely the goods are to sell out soon. This helps the retailer with forecasting — estimating how many dresses to produce and dispatch to a given market, which is crucial for any business.
What are other examples of data science being applied in the fashion industry?
Big fashion companies will have data on their sales not only online but in stores, too, and this data can be used to predict the demand of products, estimate stock needs and pricing. It can predict a customer's clothing sizes between many popular brands as well: By gathering data on people's measurements and what sizes they wear in each brand, a model can be built to predict what size you should be for a certain one, which is very useful since most products are returned because of fit.
Another great example of using data science for fashion is through a program called Deepomatic, which uses images of fashion bloggers and links the items to where they can be bought.
What do you think is a common misconception when it comes to your job?
A lot of people are finding it surprising that women do this job — not exactly at Lyst because we have a good balance — but in general and outside of fashion. It's something that I push for a lot; getting more woman in the tech world itself would be great. It should start with encouraging girls in school to study computer science or any science. It's okay to do that and okay to be a data scientist. It's actually more than okay, it's really cool at the moment. I read in Harvard Business Review that it's the sexiest job of the century and the best job to have in 2016.
Data science is the field of machine learning, or artificial intelligence, so it's an extremely exciting time to be part of it because it can be exploited in many ways, especially in the fashion industry. It doesn't stop there and it probably won't slow down anytime soon. It's changing the world of fashion.
This interview has been edited and condensed.