Meet Uber’s Identical Twin Data Scientists


Walking the hallways at Uber’s headquarters in San Francisco this summer, many people did a double-take on seeing Afshine and Shervine Amidi walk by. For the past three months, Uber employee Afshine was joined by his identical twin brother, a summer 2018 intern, in the office. Beyond looking alike, the brothers also both work as data scientists.

Afshine, a graduate of Ecole Centrale Paris and MIT, began working at Uber in 2017, using his data science skills to help our platform optimize the trip experience for both riders and drivers. Shervine, who also attended Ecole Centrale Paris, continued his studies at Stanford. For his internship at Uber, he is applying data science to Uber Eats, helping build personalization tools such as query understanding and smart food recommendations.

We sat down with Afshine and Shervine to find out about the journey that lead them to study data science and work at Uber, the challenge of using technologies such as facial recognition when you have an identical twin, and what it’s like to look alike.

Tell us a little about your background.

Afshine Amidi: Shervine and I are French, of Persian descent, born and raised in Paris. From the start, we both tended to like the same things and ended up practicing the same sports, such as judo, swimming, and water polo. When we were at school, our parents made sure to put us in different classes. This enabled us to develop and grow independently while still being within earshot of each other. One cool perk of having an identical twin is that you immediately feel close to your twin’s friends even when you have just met them for the first time.

Shervine Amidi: After graduating from high school, Afshine and I studied mathematics and physics in preparatory classes for two years and entered Ecole Centrale Paris, where we studied mathematics and general engineering. Then, Afshine went to MIT for his graduate studies, where his coursework and projects were centered on how to make the best use of data science tools in a business setting. As for me, I followed a slightly more theoretical path, as I am currently studying computational mathematics with a focus on data science from a more foundational perspective at Stanford.

I am also doing research in computer vision at the Stanford Vision and Learning Lab, where I investigate techniques that minimize the amount of labeled data needed for computers to tackle supervised tasks like image recognition. Solving this type of problem unlocks the door to endless possibilities, especially when dealing with real-world data, which often comes with noise and uncertainty.

What made you want to go into data science?

SA: Our general engineering background gave us strong foundations in various subjects, ranging from mathematics to physics, while also covering economics, law, and biology. Out of these subjects, the field of machine learning stood out to us, as it opened the door to tackling technical challenges in a way that was not possible before.

AA: In our undergraduate program, we were lucky to work with world-class professors who became informal mentors to us. We had the opportunity to do an extensive machine learning research project with Professors Nikos Paragios and Evangelia Zacharaki, where we applied machine learning techniques to solve key challenges in the field of biology.

Why did you decide to join Uber?

AA: Three years ago, Pierre-Dimitri Gore-Coty, Uber’s vice president for Europe, the Middle East, and Africa, gave a presentation at Ecole Centrale Paris (which is also his alma mater) where he laid out the vision that he had with Uber and how it was changing the way we think about transportation. His speech stood out in a lot of people’s minds, including ours, and from that moment on, we began to actively follow Uber’s progress into new markets. Another reason is that, apart from its game-changing products, Uber also has the reputation of being a data-driven tech company at the forefront of innovation.

SA: I loved seeing Afshine work on products that I use every day, and seeing those products evolve very quickly. As a user, it is exciting to see the progress Uber makes on a daily basis, and be able help improve the experience of millions of people around the world. There are still many exciting challenges in all of the businesses in which Uber is innovating, from ridesharing to Uber Eats delivery, and now Uber Freight, bikesharing through Jump Bikes, and urban air transportation with Uber Elevate.

What are your roles on your teams?

SA: This summer, I am working on the Uber Eats Data Science team. My goal is to better understand our dishes so that we can offer food and restaurant recommendations to eaters in a seamless and personalized manner.

AA: I have been working on the Uber Data Science team to make Uber more reliable while spending more efficiently. My day-to-day work involves a lot of cross-team and cross-functional interactions involving other data scientists, product managers, and software engineers. What I really enjoy about this role is that the variety of topics that I work on as well as the interactions that I have which make every day different from the ones before.

How do you apply your Data Science expertise at Uber?

AA: At first glance, one might think that Uber is just an app that connects drivers with riders. But behind the scenes, we put in a ton of effort to ensure that this matching is executed optimally for both parties. On one hand, we want riders to have an inexpensive and safe experience with a reliable pickup and low wait time, while giving drivers a seamless pickup experience and maximizing the amount of fares they find while on the road. To make that happen, we must use all of the magic that data science has to offer.

It is nice to see that the goals we set can be enhanced with statistics, machine learning, and optimization techniques. When I learnt some of these tools in grad school, I would not have imagined how powerful they could be when applied to challenges like the ones we have here at Uber. The models that I build focus on finding the right balance in our spend and product pricing to better inform key strategic decisions.

SA: What Afshine just mentioned primarily touches on the ridesharing part of Uber. I would say that what we do at Uber Eats is particularly challenging, as we have to optimize the happiness of not two, but three stakeholders at the same time: the restaurant-partner, delivery-partner, and eater. This represents a particularly complex equilibrium, as we have to be careful that the benefits of one does not negatively impact the other.

On my end, I focus on the eater side, where a strong understanding of the food options that users can choose from on the platform is crucial. Building ontologies and graphs of how different cuisines relate to each other gives us an important tool when presenting relevant food items and restaurants to eaters. My goal is to bring the smoothest and most personalized experience to anyone using the app by surfacing the dishes that suit their tastes best. In machine learning terms, this problem amounts to making the right link between eaters and the dishes that they like. Luckily, there was an awesome representation learning class I took at Stanford last year, where Professors Silvio Savarese and Amir Zamir emphasized how to leverage some of the techniques that are super useful to me today.

Do you have a fun anecdote about life as twins at Uber?

SA: Being twins can lead to some funny encounters. Oftentimes, people I don’t recognize smile at me and begin very casual talk, assuming that I am Afshine. When I understand what is going on, I like to see how far I can keep the conversation going while staying unnoticed.

AA: Apart from situations where people mix up the two of us and the pleasure of seeing amused faces on our way to work, one of the atypical sides of having a twin is that you can test the limits of the latest technologies out there; by that, I mean facial recognition. For instance, while our friends are usually able to tell us apart, it is funny to see that the phone access facial recognition feature on our phones can’t distinguish between us.

 

You don’t need an identical twin to work at Uber. Take a look at our careers page and see if there is a role for you!

Subscribe to our newsletter to keep up with the latest innovations from Uber Engineering.

Wayne Cunningham on LinkedinWayne Cunningham on Twitter
Wayne Cunningham

Wayne Cunningham, senior editor for Uber Tech Brand, has enjoyed a long career in technology journalism. Wayne has always covered cutting edge topics, from the early days of the Web to the threat of spyware to self-driving cars. In his spare time he writes fiction, having published two novels, and indulges in film photography.



Source link