Genie AI – AI shopping assistant

Genie, Snoonu’s in-app AI assistant, was designed to solve a key problem in digital marketplaces: users often struggle with overwhelming catalogs, irrelevant results, and slow product discovery caused by manual browsing or endless scrolling. By enabling natural, conversational search, Genie removes friction and helps users find what they need quickly and intuitively. It interprets intent in real time, targeting accurate, personalized recommendations that make shopping effortless and enjoyable. For users, this creates a faster, smarter, and more engaging experience that builds trust and satisfaction. For the business, Genie drives measurable impact by increasing conversion rates, average basket size, and long-term customer loyalty through a seamless AI-powered journey.

Genie runs on a cloud-native architecture blending AI reasoning, real-time data, and seamless mobile integration for intelligent, fast, and personalized experiences.

• Mobile: Composed multiplatform, iOS (SwiftUI) and Android with feature flags for controlled rollout.

• Backend: Marketplace microservices (catalog metadata, availability, coverage, deep-links and cart) and DS&AI services (Genie Multi-Agent Chatbot, Personalized Recommendations) connected via an API Gateway with SSE streaming.

• Cloud: Hosted on AWS, it leverages Bedrock foundational models (Anthropic Haiku & Sonnet) and Knowledge Bases for enhanced context and retrieval. Data pipelines built with Kinesis, Lambda, and S3 ensure continuous synchronization across systems. The Knowledge Base is refreshed hourly, ingesting updates from MongoDB change streams via Kinesis and Lambda to tackle coverage, pricing, and availability. MongoDB also stores the marketplace catalog and maintains session memory for contextual relevance during ongoing user interactions.

• AI Orchestration: PydanticAI Graph with router and recommender nodes for dynamic, context-aware decisions. Each node has its own agent, with tools specialized to accomplish specific goals.

Genie’s development involved multiple layers of complexity, primarily driven by the need for real-time, low-latency performance in a multi-agent AI system. Each user query triggers a complex sequence: detecting intent, surfing the agent graph, retrieving recommendations based on query intent and product specifications, validating availability and prices, gathering product images and metadata, attaching deep-links, generating natural conversation, and ranking results contextually, all within a live streaming chat. Implementing SSE over HTTP/2 for streaming responses was a first for Snoonu, requiring deep research and infrastructure changes. Additional challenges included ensuring safe rollout through feature flags and staged experimentation, as well as establishing strong governance for moderation, compliance, and preventing sensitive data exposure in open conversational flows.

Objectives

Genie was built to redefine the marketplace shopping experience by reducing time-to-product and increasing conversion through conversational, intent-based discovery. The goal was to enable users to find what they need naturally, without manual browsing or endless scrolling, while giving Snoonu a new competitive edge through AI-driven personalization. The project also emphasized safe deployment, measurable experimentation, and continuous iteration, shipping behind feature flags to track adoption, performance, and impact in real time.

Implementation Stages

Discovery: Defined the UX vision for “AI that shops for you,” aligning product design, customer journey, and backend research for chatbot streaming, AI modeling, and agentic decision flows.

Architecture: Focused on aligning Snoonu’s marketplace backend and mobile frontend with the DS&AI ecosystem. This phase established unified data flows, streaming APIs contracts, and event schemas to integrate personalization, catalog intelligence, and content moderation seamlessly. The result was a cohesive architecture across teams.

Mobile Integration: Built native iOS (SwiftUI) and Android components, including chat view, add-to-cart, and deep-link actions, and sessions kept.

Data & AI: Focused on AI and multi-agent development for intent understanding, business constraints, and recommendations driven by query intent and product specifications, ensuring accurate, context-aware results.

Rollout: Gradual release via feature flags, internal testing, KPI tracking, gathering customer feedback.

Post-Launch: Continuous monitoring and gathering customer feedback, prioritization of new features and improvements, leveraging analytics to refine prompts, UX, and other logic. Planning future versions.

Technology Stack

• Mobile: Composed multiplatform, iOS (SwiftUI) and Android with feature flags for controlled rollout.

• Backend: Marketplace microservices (catalog metadata, availability, coverage, deep-links and cart) and DS&AI services (Genie Multi-Agent Chatbot, Personalized Recommendations) connected via an API Gateway with SSE streaming.

• Cloud: Hosted on AWS, it leverages Bedrock foundational models (Anthropic Haiku & Sonnet) and Knowledge Bases for enhanced context and retrieval. Data pipelines built with Kinesis, Lambda, and S3 ensure continuous synchronization across systems. The Knowledge Base is refreshed hourly, ingesting updates from MongoDB change streams via Kinesis and Lambda to tackle coverage, pricing, and availability. MongoDB also stores the marketplace catalog and maintains session memory for contextual relevance during ongoing user interactions.

• AI Orchestration: PydanticAI Graph with router and recommender nodes for dynamic, context-aware decisions.

• DevOps: GitHub CI/CD, Docker, Kubernetes, AWS SAM, CloudWatch, and Datadog for scalable, observable, and reliable operations.

Business Impact

Genie transforms the user journey by making product discovery faster, context and intention aware, and more engaging. It reduces null and long search sessions while increasing add-to-cart actions and conversion rates. From a strategic perspective, it introduced Qatar’s first in-app AI shopping assistant, setting a new benchmark for digital commerce innovation in the region. By connecting AI reasoning with real-time catalog intelligence and seamless UI flows, Genie delivers measurable growth and a foundation for scalable, personalized user experiences across Snoonu’s marketplace.