Rollout to 15,000+ GitHub repositories in ~70 countries
4,000+ developers and data scientists now using Gemini Code Assist
Increased satisfaction with quality and speed of code reviews
AI inspires more collaborative culture
DORA metrics tailored for weekly snapshots of KPIs
AI-enabled coding yields faster new feature releases, improved platform stability, and better developer UX.
AI-enabled coding yields faster new feature releases, improved platform stability, and better developer UX.
The best developer tools for the best customer experience
Hankering for tapas in Madrid? Not enough time to get to o supermercado in Buenos Aires? Forget to buy flowers for the surprise party you’re throwing for Appa’s sixty-fifth birthday in Seoul? Then call Delivery Hero, a local delivery platform serving customers in nearly 70 countries on four continents, and it’ll be there in less than an hour.
Operating a local delivery platform on a global scale is no mean feat. Just ask Mert Aydin, principal software engineer at Delivery Hero, who oversees the optimization of software engineering and data science practices.
“Successfully connecting millions of customers with vendors and delivery people requires constant technological innovation,” Aydin explains. “We want to give our developers and data scientists top-of-the-line tools because a better user experience for them translates directly into the best user experience for our customers.” That means customers can find and order what they want more easily, and receive their orders more quickly.
We give our developers and data scientists top-of-the-line tools because a better user experience for them translates to the best user experience for our customers.
Mert Aydin
Principal Software Engineer, Delivery Hero
So Delivery Hero added Gemini Code Assist to its software development toolbox to address critical challenges in the company’s code review workflows. Chief among them was how pull requests were handled and the errors that resulted.
Generating pull requests and assigning reviewers had always been handled manually. That added friction to the coding lifecycle, impeded new feature delivery, and — most critically — allowed bugs to escape into software releases. In addition, vaguely written pull requests left reviewers puzzling out information about configuration, design, and library changes, further bogging down the development cycle.
“Even the most highly skilled engineers miss subtle issues like null or dangling pointers, typos, and naming inconsistencies,” Aydin acknowledges, “but subtle errors can have an outsize impact on platform reliability and customer experience, like ordering deliveries.” Aydin knew AI-enabled tools would excel in exactly those areas that needed improvement.
AI “hero” improves code reviews and accuracy
Delivery Hero already relied on Google Cloud for data analytics, and its longstanding relationship with Google was a key factor in the decision to become an early adopter of Code Assist.
For the multi-week pilot, Delivery Hero put together a diverse, cross-market cohort spanning more than 20 GitHub repositories to get the broadest understanding possible of the tool’s potential impact. “The feedback from our technical teams was immediate and overwhelmingly positive, so we chose to move forward with a full-scale rollout,” Aydin shares.
The Gemini Code Assist GitHub app is now enabled worldwide on more than 15,000 GitHub code repositories across major global and regional brands.
Developers have been especially impressed by Code Assist’s auto-summary feature, which creates comprehensive, well-documented pull requests, and the way the tool’s AI-generated comments clearly explain coding decisions, thereby accelerating the development process.
Gemini Code Assist gives developers and data scientists more time to focus on high-level issues like complex architecture and business logic evaluation — the things humans are best at. The feedback from our technical teams has been overwhelmingly positive.
Mert Aydin
Principal Software Engineer, Delivery Hero
Code Assist also suggests best practices for pull requests and flags edge cases and security flaws so developers can address them successfully earlier in the workflow. This mitigates technical debt and the need for bug fixes, leaving managers with more time for higher-value tasks, such as mentoring less-experienced developers.
The improvement in code quality and the speed of code reviews has had the biggest impact on developers. “User satisfaction scores have been uniformly positive, and adoption rates high,” Aydin reports.
Delivery Hero now uses Code Assist for initial code reviews to catch stylistic issues, typos, and other common errors.
In addition, its contextual, cross-file awareness enables it to identify where missing tests are required even when they’re triggered by changes in another file.
“Code Assist gives developers and data scientists more time to focus on high-level issues like complex architecture and business logic evaluation — the things humans are best at,” Aydin observes.
AI that helps humans collaborate more closely
Aydin has now made Gemini Code Assist available to Delivery Hero’s entire technical staff, and it’s in use by more than 4,000 software engineers and data scientists. They utilize it all across the integrated development environment (IDE) — for code completion, explanation, and unit test generation — using its browser-based interface or plugins for IntelliJ and Visual Code Studio. “Code Assist is accelerating software development before the review stage by helping developers write more consistent code from the very start,” Aydin relates.
The project has proved so successful that Delivery Hero won a 2025 Google DORA Award for improving the developer experience with Code Assist. DORA metrics are now being tailored to take weekly snapshots of key performance indicators.
Code Assist has also helped team members form new connections and work more collaboratively: Pull request authors are responding to the AI’s comments before resolving them, ensuring the entire team understands why changes have been made and making reviews more personal and interactive.
And Code Assist has complemented Delivery Hero’s own coding philosophies rather than imposing its own. By fine-tuning the AI’s suggestions with custom style guides, developers ensure the code is not only technically correct but also consistent with the company’s standards.
Delivery Hero’s goal is to make Code Assist an integral and permanent part of its software development process. The company is now exploring how Code Assist can be used to improve brownfields by explaining complex legacy code, and to accelerate onboarding by providing instant context for existing codebases. And as an early adopter, it’s been providing feedback to the Google team that’s helping to shape the product’s evolution.
“Gemini Code Assist has significantly boosted our code quality and the satisfaction of our technical teams,” Aydin concludes. “Better quality code allows us to build, test, and ship innovative new features faster and with greater confidence, enhancing the experience for millions of users as well as the stability of our platform.”
Gemini Code Assist is enabling us to ship innovative new features faster and with greater confidence, enhancing the user experience for both developers and customers, as well as the stability of our platform.
Mert Aydin
Principal Software Engineer, Delivery Hero
Get started:
Delivery Hero works with global brands and local restaurants and shops to deliver everything from meals and groceries to flowers, coffee, and medicine to customers in nearly 70 countries on four continents.
Industry: Technology
Location: Germany
Product: Gemini Code Assist
