Hotels.com CTO Thierry Bedos discusses cloud, data, DevOps and developing the diversity of the technology function

Hotels.com CTO Thierry Bedos has witnessed enormous changes to the travel industry in his eight years at the company, which have helped parent company Expedia expand to almost 90 countries, more than 60 million app downloads and revenues that have increased from $3.3 billion to $11.2 billion.

The global tech team that Bedos leads from his London office has played a key role in this growth. Bedos joined Hotels.com as director of development in 2010. At the time, the company Hotels.com was releasing once every month or two and running its workloads out of its own data centre. Today, Bedos is CTO and Hotels.com is making thousands of releases year and has gone all-in on the cloud.

Expedia has migrated the majority of its workload to Amazon Web Services, which has provided Bedos with the agility to scale services as required, break down the barriers between infrastructure engineers and developers and standardise machine learning features.

“Now we have all these teams working together to be a lot more efficient at delivering software,” Bedos tells CIO UK. “But to me, one of the main things that we are seeing from the cloud is this ability to operate data and machine learning at scale.

“If you’re not able to have large clusters to train, you’re not going to really crunch the data and be able to drive a huge amount of computing in a short period of time, and then all the machine learning algorithms and features are just not possible. It has been very transformational, for that pinpoint ability to spin up those large clusters and be able to train algorithms and a lot more efficient, and then deliver that experience to the customers.”

Developments in data

Machine learning now powers much of the customer experience at Hotels.com. When users start typing the name of a destination, algorithms suggest accommodation options based on what they’ve learned from the customer’s behaviour.

Once the destination is selected, machine learning is used to choose the best order of images for each hotel so that travellers can quickly get an understanding of each place before moving on to look at the next option, while a recommendation carousel suggests alternatives with similar features. Finally, when the customer is about to book, algorithms assess the risk of fraud and flag any suspicious activity.

Bedos believes that the next step will take the industry back to the more personal booking experience of the past.

“When you were going to your travel agents across the street, they got to know you,” he says. “They knew your preferences, they knew what you liked and didn’t like. You had a conversation with them and they could solve your problem.

“And that’s what we want to bring back to this travel experience online: more personalisation and understanding the customer better so we can serve them better and solve their problems without having them to pick up the phone and speak to an agent. How can we anticipate some of these problems and solve these problems on behalf of the customer?”

Devops for continuous delivery

Bedos has embraced devops to ensure that the online pureplay can consistently improve the customer experience by building a continuous delivery pipeline that brings together the teams that build the infrastructure and code the software.

“In the past, once developers had built their software, they would just send it to somebody else and work on the next feature. Now, they really have the accountability end-to-end of how the software is behaving on the infrastructure.

“I think that’s really necessary for us to do. Years ago, you had a number of the more monolithic types of applications. These days, with microservices, you are likely to have hundreds of services that are necessary to serve a web page, and therefore, people in the teams that build a service are also the teams that are monitoring the service and troubleshooting the service.

“For us, it’s been a shift of mentality, from a big separation between the different teams to having now these really cross-functional teams where the engineers understand everything from the architecture all the way to how to troubleshoot and monitor software online.”

Diverse requirements

Bedos wants to integrate his workforce further by embracing an inner source mentality, which applies the open source approach to internal operations. He is also working to add more diverse skills to his team is by bringing more women into the company.

“There’s lots of evidence that teams that are more balanced and more diverse in nature perform much better,” he says. “They bring better innovation to the customers out there. When you think about the fact that something like 90% of travel decisions are influenced or made by women, we need to bring more women to our teams to understand that and to design products that women will love and that will be useful for them.

To add this perspective, Hotels.com is working with schools in the UK and abroad to raise awareness of science and technology careers among girls, using apprenticeship schemes to develop talent from other sectors, and attending conferences such as Women of Silicon Roundabout to find new recruits.

“There will be a shortfall of talent in the next few years,” says Bedos. “And we are tapping into only half of the population when there is another half that’s all women that are great and have lots of talent. And so the question is, how can we bring more women to this industry.”