'The Full Lifecycle Of A Machine Learning Project Is Not Necessarily Well Understood And That Can Drive Disillusion'

Interview with Jérôme Selles, Director of Data Science at Turo


Turo, founded in 2009 under the name RelayRides until it changed its name in 2015, has become one of the shining lights of the sharing economy. It is, essentially, Airbnb for private cars, allowing owners to rent out their vehicles via an online and mobile interface, in the same way Airbnb allows homeowners to rent out their properties. It provides a far cheaper - 35% according to the company website - alternative to traditional rental companies.

The platform now has more than 2 million screened users and 110,000 vehicle listings in 4,500 cities across the US and Canada, and is now launching in the UK. It has raised $101 million to date, with the most recent Series C funding round in November 2015 bringing in $47 million. Forbes also included it in its 14 ‘hottest on-demand startups’ in 2015.

Central to its success has been its ability to exploit the data. Jérôme Selles leads data-related initiatives at Turo. From data-driven strategic decision-making to design and implementation of user-facing data features like dynamic pricing or search ranking, data is widely used at Turo and one of the key contributing factors to 4 years of hyper-growth for the company. Jérôme joined as the first data scientist in the company and has built a 15-person team. Prior to Turo, he worked as a data scientist at AgilOne, a marketing analytics startup and at SAP Labs in Palo Alto.

We sat down with him ahead of his presentation at the Machine Learning Innovation Summit, taking place this June 5-6 at the Marriott Union Square in San Francisco.

Where do you think machine learning’s most important applications will be in the near future?

I can't wait to see applications of Machine Learning completely disrupting our approach to the biggest challenges of humanity. Managing our resources, improving health, preventing risk... there are countless opportunities where Machine Learning carefully applied can lead us to a better world.

In the mobility industry, in particular, autonomous vehicles come to mind. Let's not forget that vehicles are resources that are completely underutilized today: in the US alone, there are 300 million cars and 200 million people able to drive them, in the world we're talking about 1 billion cars. Our mission at Turo is to put the world's billion cars to better use, which can be seen as a macro machine learning problem.

In a recent KDnuggets poll, 51% of respondents said that they expect most expert-level Predictive Analytics/Data Science tasks currently done by human Data Scientists to be automated to happen within the next decade. Do you think the data scientist’s role is really under threat? What kind of impact do you think it will have on the job market in general, and are governments prepared?

I think this is a great thing for the field. More and more companies need data science to help them make sense of their data. There's a real lack of people qualified for these jobs and the market is very competitive. In the last five years, we've seen the ecosystem evolve a lot and there are more tools available to handle data and learn from it, but there's still a long way to go. With more automation of these tasks, we can have human brains focus where they can have the most impact, which is more exciting and way more rewarding!

Are there any new technologies or ideas in the machine learning space that you find particularly exciting or believe will be especially important in the next few years?

I'm pretty excited to see Google getting into the field as a vendor and providing their expertise in the domain with the different parts of Google Cloud. Also, the understanding of the processes and the infrastructure it takes to make a real world application of machine learning is getting better and better. It is great to see progress being made in abstracting the infrastructure complexity and making it easier for more data-savvy people to manipulate datasets. Amazon has done an amazing job in that space, but it's especially exciting to see Google becoming a challenger. Along with that, more and more projects are becoming open source - this emulation should elevate the ecosystem as a whole.

What challenges do you foresee holding back machine learning from achieving its potential? How do you think these could be overcome?

When it comes to applications of machine learning, the expectations are usually very high. The full lifecycle of a machine learning project is not necessarily well understood and that can drive disillusion within an organization and for the users. Depending on the quality of the data that is being used, automating the learning loop can be a challenge and, today, requires manual supervision. A good illustration of that is what happened with the Microsoft chatbot Tay that became racist within 24 hours. For Machine Learning to achieve its own potential, the learning process needs to be kept under control and values need to be respected. Data quality for the models is as important as education values in our society and we need more automated and systematic ways to make this happen.

What will you be discussing in your presentation?

I will talk about the implementation of Machine Learning at Turo for ranking relevance. With tens of thousands of cars available in our main markets, it becomes crucial for us to show our travelers the most relevant cars for their trip. From data collection to modeling and all the way to implementation and experimentation, I will talk about the different steps of the lifecycle of the machine learning model that powers today our ranking at Turo.

You can hear more from Jérôme, along with other leading experts in the field from the likes of Google, Amazon, and Facebook, at the Machine Learning Innovation Summit. View the full agenda here.

BONUS CONTENT: Teymur Sadikhov, Senior Vehicle Intelligence Engineer at Mercedes-Benz USA, discusses how machine learning is contributing to the revolution in autonomous vehicles in the areas ranging from perception to planning



comments powered byDisqus
Data culture small

Read next:

Building A Culture Of Data