Uber vs. Lyft: How the rivals approach cloud, AI, and machine learning

Uber is launching its IPO at $45 a share and Lyft is already public. Here's a look at how they approach technology, infrastructure and development.
Written by Larry Dignan, Contributor
Uber, Lyft

Uber and Lyft are roughly in the same business and aiming to reinvent transportation, but their technology approaches have more differences than similarities. Sure the technology strategies rhyme in places, but Uber's approach is broader as it eyes multiple businesses.

When comparing the regulatory filings of Uber and Lyft some technology differences become clear. First, Uber sees itself as building a marketplace and technology platform that can extend into multiple areas (Uber Eats and Uber Freight for instance) while Lyft sees itself as primarily in transportation as a service provider.

Uber, which is kicking off its initial public offering priced at $45 a share, also touts that it has "a team of more than 3,000 highly skilled engineers and computer scientists whose expertise spans a broad range of technical areas" and an automated infrastructure. Uber has 22,263 global employees. Lyft said 36 percent of its 4,791 employees work in its product management, engineering and design organizations.

Meanwhile, Uber reported 2018 net income of $987 million on revenue of $11.27 billion. Lyft reported a 2018 net loss of $911.3 million on revenue of $2.16 billion.

Here's a look at the approaches on key technologies by Uber and Lyft.


Uber appears to have a classic hybrid cloud approach. Uber has co-located facilities and multiple cloud vendors. Uber said:

We have developed our infrastructure to be highly automated, enabling us to improve our platform and add new features with rapid velocity. We built our platform to handle spikes in usage, such as those we experience during holidays. We currently use multiple third-party cloud computing services and have co-located data centers located in the United States and abroad. These partnerships allow us to quickly and efficiently scale up our services to meet spikes in usage without upfront infrastructure costs, allowing us to maintain our focus on building great products.

Uber also said that it has commitments for network and cloud services as well as background checks with varying expiration terms through 2020. Uber's filing also includes order forms for upgrading access to Google Maps APIs.

Lyft has bet on Amazon Web Services for its architecture and has agreed to spend at least $300 million between January 2019 and December 2021. Lyft said the AWS partnership allows it to be more resilient to surges, but said if it fails to hit its minimum purchase requirement with AWS it may be required to pay the difference.

"We pay AWS monthly, and we may pay more than the minimum purchase commitment to AWS based on usage," said Lyft. AWS may only terminate its Lyft agreement for convenience after March 31, 2022 with advance notice.

Must read

Artificial intelligence and machine learning

Uber mentions the term artificial intelligence six times in its IPO filing, highlights machine learning 11 times and the word algorithm 16 times. Those mentions aren't by accident. Uber makes it clear that its data science and algorithms are the key to its marketplace technologies.

The company said:

We have made substantial investments in AI and machine learning. We have created and grown a world-class research team that has produced numerous original publications, patented technologies, and widely-used open source software. Managing the complexity of our massive network and harnessing the data from over 10 billion trips exceeds human capability, so we use machine learning and artificial intelligence, trained on historical transactions, to help automate marketplace decisions.

Uber said it has built a machine learning platform as well as natural language and dialog system technologies. Computer vision "automatically processes and verifies millions of business-critical images and documents such as drivers' licenses and restaurant menus, among other items, per year."

In addition, Uber said it uses algorithms for sensor processing for location accuracy, crash detection as well as matching drivers and passengers. Routing technologies are also based on algorithms that handle thousands of ETA requests per second.

Also: What is deep learning? Everything you need to know | What is machine learning? Everything you need to know

Lyft makes no mention of artificial intelligence in its IPO filings, but does note miles traveled and rides "inform our machine learning algorithms and data science engines." The company also said:

We leverage insights from this data to improve the product experience for riders by presenting them with personalized transportation options. The more rides we facilitate, the better we are able to improve our matching efficiency between drivers and riders in our ridesharing marketplace, which reduces arrival times and maximizes availability to riders. Our data insights also allow us to anticipate market-specific demand, enabling us to create customized incentives for drivers in local markets. We enable riders to optimize routes across multiple modes of transportation which we believe provides us with a significant advantage over single modality providers.

Algorithms also inform driver incentives, dispatching and availability and pricing.

Must read

Marketplace technology

Uber has built its own proprietary marketing, routing and payment technologies. The bridge between Uber's various ventures is its marketplace technologies. Uber said:

Our marketplace technologies comprise the real-time algorithmic decision engine that matches supply and demand for our Personal Mobility, Uber Eats, and Uber Freight offerings.

The marketplace tools include:

  • A demand prediction engine to predict volume, supply and demand and location dynamics with current and historical trends. Data visualizations are available for zones with unique pricing characteristics.
  • Matching and dispatching algorithms that review and consider variables such as distance, time, traffic, weather and even meal preparation times for Uber Eats.
  • Pricing tools that set real time prices at the local level based on demand.

Lyft said that its platform leverages historical data, supply and demand as well as driver availability. "Utilizing machine learning capabilities to predict future behavior based on many years of historical data and use cases, we employ various levers to balance supply and demand in the marketplace, creating increased driver earnings while maintaining strong service levels for riders. We also leverage our data science and algorithms to inform our product development, such as the introduction of our current subscription product," the company said.

Marketplace efficiency is powered by Lyft's dispatch platform and data such as distance, route, traffic and travel. Lyft prods drivers to go in areas based on sustainable prices so they can increase earnings.

Must read


Uber said its development approach revolves around agile and DevOps. Uber said:

We are constantly prototyping, experimenting, launching, and refining our products to deliver the best experience to tens of millions of platform users. We conduct staged rollouts when testing new products and features, often initially deploying to a small portion of platform users, such as a single neighborhood or city district, to gather feedback, monitor performance, and course correct as necessary. The size of our network enables us to introduce new features and observe performance at a speed, efficiency, and scale that we believe our competitors cannot match. We then gradually scale these products and features to reach additional platform users, while continuously optimizing performance throughout.

Uber also said its plan is to integrate its key technologies. For instance, Uber's payment platform is integrated into its technology stack and is enhanced by payment partners.

TechRepublic: 7 must-have skills for enterprise DevOps practitioners | In-demand programming language Java scores another win as DevOps turn to its Groovy spin-off | How to avoid the top 5 causes of DevOps failure

Lyft's model revolves around organizing product teams with a development model that integrates product, engineering, analytics, data science and design. Lyft also relies on third party open source software. 

According to Lyft, it follows a continuous deployment strategy. "We rely heavily on a software engineering practice known as "continuous deployment," which refers to the frequent release of our software code, sometimes multiple times per day," the company said. 

Autonomous vehicles

Both companies are betting that autonomous vehicles will power their transportation networks over time. The Uber and Lyft approaches, however, differ.

Uber noted that its Advanced Technologies Group is focused on autonomous and has more than 1,000 employees dedicated to next gen transportation. The company in its IPO filing said that it thinks there will be "a long period of hybrid autonomy" where duties are split by the vehicle and drivers.  As a result Uber is partnering with the likes of Toyota and Volvo to integrate technologies.

The company is also exploring autonomous technologies via aerial ride sharing through Uber Elevate between suburbs and cities.

Lyft touted its open autonomous vehicle platform to connect its network with partners. The company said:

Through our Open Platform, we enable partners to connect with our network and offer their autonomous vehicles on the Lyft network. For example, our Open Platform partnership with Aptiv has enabled the commercial deployment of a fleet of autonomous vehicles on our platform in Las Vegas. We have facilitated over 35,000 rides in Aptiv autonomous vehicles with a safety driver since January 2018.

Partnerships are one prong of the autonomous strategy. While it has an open platform, Lyft also said it is "building our own world-class autonomous vehicle system at our Level 5 Engineering Center." The bet is that the combination of autonomous partnerships and its own intellectual property will be an advantage.

Must read

What's in a name? These DevOps tools come with strange backstories

Editorial standards