AWS Partner Network (APN) Blog
Building intelligent data foundations: The Grubhub and Vody partnership on AWS
By: Dr. Yubin Kim, Chief Scientist – Vody
By: Nneoma Okoroafor, Partner Solutions Architect – AWS
By: Swagat Kulkarni, Sr. Solutions Architect – AWS
By: Siraj Gadne, Sr. Mgr, CSM, US SWIM – AWS
 |
| Vody |
 |
This post was co-written with Grubhub and Vody, an AWS Partner, for data engineers, ML platform teams, and retail technology leaders. Familiarity with Amazon EKS, Amazon Bedrock, and data pipeline concepts is helpful but not required.
Retailers are racing to adopt AI, but the companies that pull ahead won’t win because they shipped a single breakthrough feature. They’ll win because they invested early in the data foundations that can make AI initiatives faster, smarter, and more effective. At Amazon Web Services (AWS), we see this pattern across the most innovative customers: the shift from fragmented data projects to unified, intelligent data layers built on top of cloud-centered infrastructure.
Grubhub and Vody are taking that shift directly. Together, they’re modernizing the core building blocks of catalog, content, and behavioral data using AWS services and building the durable foundation required for the next era of AI-powered commerce. The early results are promising:
- More than 400 thousand merchants and tens of millions of menu items cataloged and updated regularly; ensuring fresh, structured data flows reliably into downstream systems
- Delivered results from initial model deployment in only 6-10 weeks, proving that major retailers can move quickly when the foundation is right
- Improved search recall, with initial tests showing improved relevance of candidates retrieved
- Higher conversion and click-through rates for dietary-intent searches, showing how enriched data directly improves customer experience
What’s emerging isn’t just better search and recommendations. It’s a future where structured, AI-ready data enables personalization that feels intuitive, discovery that understands intent, and merchandising that keeps pace with fast-moving cultural and consumer trends. Grubhub and Vody’s work on AWS is a blueprint for retailers preparing for what comes next: a world where the companies with the strongest data foundations will lead every wave of AI innovation.
Why the data foundation matters now
Three forces are rewriting how retailers think about data:
Generative AI everywhere: Large language models (LLMs) can analyze, recommend, and create, but they perform best when sitting on top of structured, consistent, buyer-relevant data. AI can’t invent signals retailers never stored. AWS customers across commerce, marketplaces, and delivery are recognizing this opportunity and investing in their data foundations.
Demand signals live outside the organization: Trends start on TikTok, Instagram, and community spaces long before they show up in dashboards. Modern catalog enrichment workflows need to operate at this velocity-updating in hours, not quarters, and syncing internal datasets with emerging external demand.
The opportunity for unified data: When search teams, merchandising, CRM, and marketing build on a shared source of truth, experimentation accelerates and AI projects scale more effectively across functions. Retailers are moving toward unified data platforms that all systems and teams can build on.
Establishing the foundation: Clean, structured, buyer-relevant data
Grubhub, with more than 400,000 merchants and tens of millions of menu items, recognized the opportunity to build a data layer that could support AI workloads at scale. Built on AWS, Vody transforms massive volumes of unstructured merchant content into clean, structured, AI-ready data.
Figure 1: Vody AWS architecture
The architecture, shown in Figure 1, implements network segmentation with public and private subnets. Customer traffic enters through an application load balancer, while outbound connectivity from the Amazon Elastic Kubernetes Service (Amazon EKS) cluster (such as calls to Amazon Bedrock) routes through a NAT gateway, maintaining security isolation for compute resources.
The platform uses Amazon EKS for container orchestration and integrates with Amazon Bedrock for generative AI capabilities. Amazon Relational Database Service (Amazon RDS) provides relational data storage, while Amazon ElastiCache serves dual purposes: as a queuing backend for background job processing and as a caching layer for API endpoints, optimizing overall system performance. AWS IAM Identity Center provides secure access management across the platform.
Using the diverse model portfolio provided by Amazon Bedrock (Claude, Nova, and Llama families), Vody generates customer-aligned search queries that enrich even low-data SKUs, ensuring consistent discoverability across the entire catalog.
The Vody platform operates as a software as a service (SaaS) solution hosted by Vody, providing merchants with a fully managed catalog enrichment service. With this managed approach, retailers can benefit from continuous AI improvements and scale without infrastructure overhead.
The platform delivers:
- Normalized, enriched product and item attributes aligned to real-world buyer behavior
- Labels that reflect how people discover restaurants and food
- Structured metadata that feeds both traditional ranking systems and modern generative models
- Frequent updates as restaurants, menus, and items change
This work represents the first key milestone toward establishing a future-looking data foundation at Grubhub and delivers measurable value from day one: better search recall, higher search conversion, and more intuitive product discovery-all without replacing existing search engines.
“Developing this in-house may have shifted focus away from our core product. After evaluating the resources required, we decided a third-party solution was the more efficient path”
– Jayesh Kataria, Sr. Director of Engineering, Grubhub
Results: Measurable impact across the business
The AWS-powered Vody platform now processes merchant data at impressive scale:
- Daily processing: Over 500,000 items enriched daily
- Weekly throughput: 2-5 million items processed
Business impact and ROI
The 12-week deployment to full production demonstrates that major retailers can move quickly when building on the right foundation. Key business outcomes include:
- Enhanced search recall: More candidate merchants retrieved and more results matching actual customer intent during preliminary analyses
- Higher conversion rates: Particularly for dietary and food characteristic searches
- Faster time-to-market: Teams can now ship campaigns, taxonomies, and product updates in days instead of weeks
- Optimized resource allocation: Strategic investment in data infrastructure versus building custom solutions
“We’re running enrichment across millions of items in our catalog across multiple facets, effectively reshaping and simplifying how our search and discovery systems operate”
– Jayesh Kataria, Sr. Director of Engineering, Grubhub
Towards an intelligent data platform on AWS
Consistent, structured product data is just the beginning. Vody envisions a roadmap toward what we call an intelligent data platform (IDP): a cloud-centered, AI-ready layer that unifies signals and supports every downstream experience.
An IDP on AWS brings three capabilities together:
- Unified signals: Catalog, clickstream, availability, and fulfillment data live alongside external trends, search behavior, and micro-signals from social platforms. AWS services such as Amazon Simple Storage Service (Amazon S3), AWS Glue, Amazon OpenSearch Service, and Amazon Redshift make this unification possible at scale.
- Adaptive metadata: As buyer language shifts, attributes evolve without breaking pipelines. Merchants and brands can introduce new concepts-aesthetics, dietary signals, styles, emerging themes-as needed.
- Shared context and memory: Every A/B test, campaign, and model output feeds back into a shared knowledge layer. Teams stop re-creating their own one-off datasets and instead build on a common foundation. This is how retailers move from large amounts of unstructured data to a true AI data strategy.
After an IDP is in place, the business impact compounds:
- Better search and discovery: Search engines unlock new relevance opportunities. Queries map to buyer intent, not just keywords.
- Personalization that feels human: LLMs and ranking models operate on rich contextual signals rather than sparse metadata.
- Faster experimentation: Teams ship campaigns, taxonomies, and product updates in days instead of weeks.
- New AI surfaces: A structured foundation on AWS unlocks next-generation acquisition and engagement channels-from Generative Engine Optimization (GEO) to conversational shopping to agent-driven merchandising.
Retailers choose AWS for this work because AWS services provide:
- Scalability: Ingestion of millions of SKUs, items, documents, and user behaviors
- Security and governance: Essential for enterprise-grade catalog and customer data
- AI/ML integration: Seamless use of Amazon Bedrock, Amazon SageMaker, and OpenSearch for retrieval and inference
Grubhub’s deployment demonstrates how quickly retailers can reach scale. Within 6-10 weeks, the joint AWS-Vody solution enriched millions of items, and pushed cleaner structured data into downstream systems while keeping merchant updates current, creating enormous potential to lift search performance.
The evolution of this work
Building an intelligent data foundation isn’t a one-time effort, it’s an iterative process. With the foundation in place, the team focused on scaling, speed, and operational efficiency. Three areas stood out:
- Scaling across merchant diversity: The team built modular extraction pipelines that could adapt to merchant variability without retraining entire models, enabling consistent attribute structure across hundreds of thousands of merchants.
- Rapid integration impact: Enriched attributes began influencing downstream relevance signals quickly once integrated. Improvements were measurable without requiring changes to the core ranking infrastructure.
- Continuous optimization: Early alignment on evaluation metrics and performance dashboards streamlined cross-team visibility and reduced iteration cycles during rollout.
Looking ahead: The shift to intelligent retail
Many retailers are somewhere on the journey from fragmented data to intelligent data foundations. Grubhub’s work with Vody on AWS shows what the next stage looks like: a flexible, adaptive data layer that keeps pace with consumers, channels, and AI systems.
The message is clear: The future of retail belongs to companies that invest in their data foundation today. After the structure is in place, new AI initiatives can be implemented faster, with less effort, and be more valuable.
AWS is committed to helping retailers make that move and to supporting partners like Vody, who are building the connections that enable the AI future of commerce.
Conclusion
The Grubhub-Vody partnership demonstrates that AI-ready data foundations deliver real outcomes fast: tens of millions of entities enriched monthly, all within 6-10 weeks on AWS-accomplished by investing in a unified data infrastructure at the start. AI initiatives that follow can be faster and more effective.
Ready to build your intelligent data foundation?
Vody – AWS Partner Spotlight
Vody is an AWS Advanced Technology Partner that transforms messy, static product catalogs into enriched, continuously updated data feeds optimized for downstream AI performance, including on-site search, generative engine optimization, advertising, and emerging agentic shopping assistants.