Business Problem

SeededHome aims to revolutionize how users buy furniture by providing a hyper-personalized shopping experience. The traditional furniture buying process can be stressful and time-consuming, with users often needing help finding the right products that match their preferences. Understanding users’ shopping preferences and profiling their personas based on these preferences is crucial to simplifying and enhancing the furniture buying journey. SeededHome identified several key challenges: the complex buying journey where users face a lengthy and complicated process to find suitable furniture, the stress and uncertainty in determining the right products, and the lack of personalization as existing platforms do not offer tailored recommendations based on individual needs. These challenges significantly affect user satisfaction and operational efficiency, highlighting the need for a more streamlined, efficient, personalized shopping experience.

About Seeded Home

SeededHome is an online platform that simplifies home furnishing. They offer a curated selection of established and emerging homeware brands, personalized based on your taste and lifestyle. Frustrated by impersonal shopping experiences, the founders created SeededHome to make finding furniture and decor fun and stress-free. With on-demand design consultations and a personal concierge, SeededHome helps you create a home you love.

Solution

SeededHome is developing a personalized shopping assistant powered by Generative AI technologies to address these challenges. This solution aims to provide a seamless, efficient, and highly customized shopping experience. GoML proposes building a product recommendation and ranking engine using the AWS technology stack and Lyzr SDKs. Key features include:

Immersive Shopping Experience

The shopping assistant will offer an immersive experience, allowing users to select their preferences and create a detailed shopping persona. This persona will help in tailoring the recommendations to individual tastes and requirements.

Relevant Product Outcomes

By leveraging advanced AI algorithms, the assistant will provide the most relevant product matches, significantly reducing the buying journey and enabling users to make decisions faster.

Conversational Interface

The assistant will feature an NLP-driven, free-flowing English conversational interface, making it easy for users to interact with the system, ask questions, and get immediate responses.

Architecture

• API and User Interaction: API Gateway acts as the front door for all API calls, routing them appropriately to backend services.
  PHP on EC2 serves as the existing backend system handling the primary business logic and user interactions.
• Data Processing and Management: Lambda Functions handle specific tasks like data ingestion and conversation logic, using in a serverless environment to manage execution without server overhead.
  Step Function orchestrates complex workflows for bulk data ingestion, ensuring processes are completed in sequence and handling errors effectively.
• AI and Machine Learning Operations: Bedrock – Embedding on Amazon Inf1 instances is used for high-performance machine learning inference, specifically for embedding generation.
  Bedrock – Claude in ECR suggests the use of AI models, possibly for natural language processing or other cognitive services.
• Data Storage Solutions: MongoDB on EC2 provides a flexible NoSQL database solution for storing conversational data and other non-relational data types.
  Relational Database manages structured data, suitable for transactions and complex queries that require relational integrity.
  Weaviate on EC2 acts as a vector search engine, enabling efficient indexing and retrieval of large volumes of vector data.
• Infrastructure and Monitoring: Cloud Watch offers comprehensive monitoring and observability across all AWS services, helping maintain system performance and operational health.

‍