The Product Dish: How Grubhub Built Its First Marketplace Robot Delivery Experience with Avride

To showcase the product innovation work happening across Grubhub, we are taking a closer look at the technology and product initiatives improving how customers, merchants, and partners connect through our platform.

Autonomous delivery has been part of Grubhub's campus ecosystem for several years, and in October, we launched our first non-campus autonomous delivery pilot in Jersey City with Avride, marking an important milestone in our autonomous fulfillment efforts.

Autonomous delivery has already shown strong potential in campus environments. However, its long-term impact depends on how it performs in the complexity of the broader marketplace, where most Grubhub orders take place. Expanding robot delivery into the open marketplace introduces a different set of product, technical, and operational considerations.

Campus environments tend to be more controlled, with defined geographies, predictable demand patterns, and standardized infrastructure. Marketplace delivery introduces a wider range of variables, including diverse merchant types, residential building formats, curbside conditions, and navigating dense urban areas.

With this pilot, we aimed to understand how autonomous delivery could support:

• Improved delivery efficiency in dense urban areas
• A differentiated customer experience through increased visibility and predictability
• Future autonomous delivery formats

What's new?

This pilot introduces new platform and product capabilities across the diner, merchant, and internal operations experience.

1. Robot delivery as a first-class fulfillment option: Robot delivery appears as a standard option at checkout on Grubhub for all orders from Wonder's Jersey City location. After checkout, our system evaluates real-time eligibility based on delivery address, robot availability, and service conditions. If a robot is available, the order is fulfilled autonomously. If not, it seamlessly falls back to a traditional courier.
2. Real-time autonomous tracking on order tracking: Once assigned, diners can track their robot's live location directly on Grubhub's order tracking page. We extended our existing courier map and status systems to support autonomous vehicle telemetry, including robot-specific delivery states.
3. Secure, app-based handoff: When the robot arrives, diners receive an in-app flow to unlock the robot, retrieve their food, and confirm closure. This enables a secure, fully self-service handoff experience.
4. Merchant-side robot fulfillment tools: Wonder staff use Avride-provided tablets for robot dispatch, loading, and readiness tracking to support smooth preparation and launch of each autonomous delivery.
5. Care enablement and edge-case flows: We established a dedicated group within our existing Customer Care organization to support autonomous delivery scenarios. This team was equipped with new workflows for robot-specific cases, including unlock issues, unresponsive diners, and robot malfunctions, extending our support systems to handle autonomy-related edge cases as we scale. Avride also escalates real-time robot and delivery issues to Grubhub through a dedicated communication channel. This allows our Care and Operations teams to coordinate and respond quickly during live incidents.

How does it work?

Robot delivery on Grubhub starts at checkout, where diners ordering from Wonder's Jersey City location see robot delivery as an option. After checkout, our backend evaluates real-time factors, such as distance, service constraints, and robot availability, to determine whether the order will be assigned to a robot or routed to a standard courier.

Once assigned, the order flows through Grubhub's fulfillment orchestration system and is dispatched via Avride's platform. Wonder staff prepare and load the order into the robot using Avride's tablet system before departure. As the robot navigates the delivery route, Avride streams real-time location and status data into Grubhub, powering live map tracking and delivery updates in the app. When the robot arrives, the diner unlocks it through the app, retrieves their food, and closes the compartment, allowing the robot to return to the restaurant for its next delivery.

How we partnered with Avride and Wonder

Launching autonomous delivery in the marketplace required close product, engineering, and operational collaboration across teams. In the weeks leading up to launch, Grubhub's Product, Engineering, and Operations teams worked closely with both Avride and Wonder through recurring cross-functional sessions focused on system design, API integration, and operational readiness.

On the technical side, teams collaborated directly on integrating Avride's dispatch and telemetry APIs into Grubhub's fulfillment orchestration and order tracking systems. This included aligning on shared status models, error handling, robot assignment logic, and fallback workflows.

Operationally, we conducted nearly four weeks of structured pre-launch testing. These sessions allowed us to run end-to-end simulations, from order placement and merchant loading to autonomous navigation and diner handoff, under real delivery conditions.

This partnership enabled faster iteration, surfaced edge cases early, and helped ensure both the technical systems and on-the-ground operations were ready before opening the experience to diners.

What's next?

This pilot will help inform how we scale autonomous delivery across Grubhub's broader marketplace. Next steps include:

• Expanding to additional markets
• Iterating based on diner, merchant, and operational feedback
• Enhancing ETA modeling, fallback handling, and concurrency management

Autonomous delivery is an important part of Grubhub's long-term approach to last-mile fulfillment. Through pilots like this one, we continue to learn, iterate, and build the systems needed to support new delivery formats while maintaining reliability and performance for diners and merchants.

Want to be part of the magic behind our product launches? Explore our open Product roles here!