At Marvik, we’ve had the opportunity to work closely with Rappi on a series of AI-driven initiatives aimed at addressing strategic challenges across the business. Our most recent joint efforts focused on three core areas: enhancing user experience, improving delivery performance and supporting restaurant business growth.

Each project was designed with a clear objective in mind:

• AI Powered Search and Recommendations: Aimed at improving the experience for end users by using machine learning and natural language processing to help them discover and find exactly what they’re looking for, faster and more intuitively.
• Agentic AI for Stores Growth: Developed to give Rappi’s merchant partners actionable insights and personalized strategies to grow their businesses, increase visibility and drive revenue.
• AI for Courier Efficiency: Focused on providing data-driven tools and improved operational guidance based on delivery feedback, to help reduce order cancellations and optimize delivery routes, improving both efficiency and user satisfaction.

By combining cutting-edge machine learning, large language models, computer vision, agents and automation, these solutions reduce manual work, provide valuable business insights and enable smarter decision-making across the Rappi ecosystem. These projects highlight how applied AI can generate scalable and measurable results when aligned with real operational needs.

The challenges

• Growth and visibility for restaurant owners: Many restaurant partners lacked visibility into their business performance and had no clear path to drive growth. Internal teams also spent significant time manually analyzing data, restricting scalability.
• Delivery efficiency: Riders often faced inaccurate store locations, inconsistent delivery instructions and time-consuming manual validations, affecting delivery speed and customer satisfaction.
• Data accessibility: Business users needed a more intuitive and user-friendly way to explore and leverage internal data, without depending on technical support.
• User experience optimization: Improving the end-user experience was essential, from reducing wait times and delivery uncertainties to providing smarter assistance during the search process, helping users find what they need more quickly and accurately.
• Reducing order cancellations and delays: Helping restaurants streamline their operations to reduce long wait times and minimize the risk of undelivered or canceled orders, improving the reliability of the platform for all users.

The solutions

• Agentic AI system for merchant growth: We built an LLM-powered system capable of understanding merchant queries, generating analysis plans, executing those analyses and delivering real-time, actionable insights. The system aims to improve key metrics such as conversion rate, average order value, customer retention and merchant support resolution time.
  • It is equipped with tools such as text-to-SQL, dynamic chart generation, tabular data analysis and conversational memory to handle follow-up questions.
  • These analyses were shown to the merchants in Rappi's website, where dynamic elements such as tables and charts were shown within chats.
• LLMs and Computer Vision models for riders operations: We apply machine learning and computer vision to:
  • Improve store location accuracy using historical rider arrival data.
  • Standardize delivery instructions using LLM-generated summaries. ○ Automate validation of rider-uploaded photos to improve the overall experience and reduce proactive compensations (e.g., store closed or proof of delivery).
• LLMs and Sequential Models for Smarter Search and Shopping Personalization: We made it easier for users to find what they need, and discover more, by applying advanced AI models like LLaMA:
  • Accelerated product discovery through smart chips that surface relevant filters and suggestions based on extracted product attributes, helping users refine their search with just a tap
  • Increased relevance of recommendations by analyzing shopping carts to detect user intent and suggesting complementary products aligned with their current mission.
  • Helped users complete their baskets with predictive models that anticipate what they’re likely to buy next, unlocking new cross-sell opportunities

Impact and benefits

• Productivity: Non-technical teams gained faster access to insights, significantly reducing repetitive tasks for internal analysts.
• Order cancellation and proactive compensation reduction: Riders spend less time navigating errors, leading to higher first-attempt delivery success rates and fewer support claims due to issues like missing proof of delivery or store closures.
• Growth Enablement: Merchant partners received consistent, data-backed recommendations to optimize ads, promotions and operations.
• Scalability: All solutions were designed on modular, maintainable architectures that can scale across teams, markets and future use cases.

Why This Matters

As on-demand platforms grow in complexity, the ability to deliver personalized, scalable and efficient experiences becomes critical. This project demonstrates how applied AI can move beyond experimentation to deliver real, sustained business impact.

By embedding AI into core operations, from search and delivery optimization to merchant support, Rappi has strengthened its agility, improved user satisfaction and enabled smarter growth strategies. It’s a clear example of how AI is not only a tool for innovation; it has become foundational infrastructure for eCommerce.

Looking ahead

We believe this is just the beginning. The same AI systems deployed today to improve operations have the potential to evolve into autonomous agents that make proactive business decisions, enhance hyper-personalization and streamline coordination across the Rappi ecosystem while maintaining brand and content standards.

With this strong foundation in place, Rappi is well positioned to continue pushing boundaries, scaling smarter, faster and more efficiently.