Summary
  • The customer operates a global wine discovery and marketplace platform, establishing a membership-driven community to unite wine enthusiasts, foster knowledge sharing, and support the curation of unique wine collections.

  • Their initiative involved creating a Conversational Agent to aid wine connoisseurs by providing information on various wines within their storage space and their distinctive traits. Moreover, the Agent is equipped to address general inquiries regarding the wine industry and its diverse wine types. The envisioned solution is designed to be integrated into the Metaverse, offering virtual consultancy services accessible via the company's VR extension.

  • The project entailed developing a system that utilizes API-driven Large Language Models (LLMs) such as OpenAI GPT-3.5, alongside custom models for question categorization, embedding creation, and named entity identification. Milvus was employed as a vector database for storing knowledge, with this tech stack combination driving the development of the search engine and enhancing the interactive capabilities of the agent.

Business Impact
  • Enhanced Customer Engagement: The AI-powered consultant provided personalized wine recommendations and expert advice, enhancing customer engagement.

  • Improved Customer Satisfaction: Quick and accurate responses to customer queries led to higher customer satisfaction and positive feedback.

  • Increased Sales: Personalized recommendations helped in converting inquiries into purchases, boosting the client’s revenue.

  • Scalability: The scalable architecture ensured the consultant could handle peak usage periods without performance degradation.

  • Customer Loyalty: The personalized experience and the consultant’s ability to store customer preferences helped in building customer loyalty.

Tech challenges

The primary technical challenge was developing a highly accurate NLP model capable of understanding diverse customer queries related to wine and providing relevant recommendations and information. Additionally, integrating the consultant with the client’s existing systems, such as their inventory and customer databases, posed significant technical difficulties. Ensuring data security and privacy while handling sensitive customer information was also a critical challenge.

Timelines
1

2 weeks

Solution Architecture Design

Deployment of DRL's Internal Knowledge Bot by Solution Architect

2

4 weeks

Data Integration

Data Integration Pipelines Development Data Cleaning & Preprocessing

3

3 weeks

Customization and Training

Agent Case-Specific Customization and Training Search Query Optimization

4

1 week

Integration with a Metaverse

Integration with a Metaverse Avatar, Testing & Deployment

Case Study Info

  • Industry:
    Retail and Technology (Wine Industry)
  • Stack:
    Python, AWS cloud stack, Deberta LM, Spacy NER, Milvus, Postgres, OpenAI API

Highlights

Case Study

See More Case Studies.