Case Reduction Chatbot

The Requirement:

Diagnose issues with Volvo or Mack trucks faster and provide with the best possible information to fix the issue. This part is usually the bottleneck when technicians serve customers.

How did we solve it?

• Built an RAG pipeline using Azure AI search and OpenAI
• Split up documents using code, to allow for token limitations
• Created search indexes for internal case reduction documents
• Used the abilities of OpenAI to summarize the findings
• Added history summarization to allow OpenAI to understand the context (at this moment, the api doesn’t hold history)
• Developed a streamlit application for users to access it
• Deployed into Volvo internal apps using Azure container services and pipelines

What did we achieve?

• A faster way for technicians to serve customers allowing for more cases to solved within the same time
• Opened up a generalized codebase, capable of being used for multiple other use cases

Technologies:

• Azure Services: Azure Storage, Pipelines, docker
• Programming Languages: Python
• ML Models: OpenAI
• Frontend: Streamlit

Team:

• Abhimanyu Bellam, Son Nguyen, Kelechi Ikegwu (Manager)