Project: Reasoning Agents - EscalationAgent-RAG #74
Description
Track
Reasoning Agents (Azure AI Foundry)
Project Name
EscalationAgent-RAG
GitHub Username
Repository URL
https://github.com/kcherur/EscalationAgent-RAG
Project Description
This project implements an Agentic (RAG) system for intelligent customer complaint escalation. It combines structured data with embedding-based semantic retrieval to perform historical pattern analysis and detect recurring issues. By integrating Azure OpenAI, Azure AI Search, and an autonomous reasoning agent, the system moves beyond traditional semantic search to enable aggregation-driven decision workflows. The result is a scalable, memory-aware escalation framework capable of identifying systemic product issues rather than isolated complaints.
Demo Video or Screenshots
Added read me and set up. I am still working on providing more details on documentation.
If time permits, I am going to enhance project and documentation.
Primary Programming Language
Python
Key Technologies Used
azure-ai-agents ( Azure AI search and Azure Open AI)
azure-ai-projects ( azure foundry)
Submission Type
Individual
Team Members
Its individual project
Submission Requirements
- My project meets the track-specific challenge requirements
- My repository includes a comprehensive README.md with setup instructions
- My code does not contain hardcoded API keys or secrets
- I have included demo materials (video or screenshots)
- My project is my own work with proper attribution for any third-party code
- I agree to the Code of Conduct
- I have read and agree to the Disclaimer
- My submission does NOT contain any confidential, proprietary, or sensitive information
- I confirm I have the rights to submit this content and grant the necessary licenses
Quick Setup Summary
follow setup commands to make run environment ready.
You must manually create the following Azure resources:
create Azure AI Foundry
Create Azure OpenAI resource
Deploy:
GPT model (e.g., gpt-4o or similar) for reasoning
Embedding model (e.g., text-embedding-3-small)
Note-down: OpenAIEndpoint, OpenAI_apiKey Deployment-names
Create Azure AI Search service
Create a Vector-enabled Index using -rag_index.json
Note-down: AzureAISearch-endpoint, Azuresearch-Admin-APIkey,Index name
add all these variables in .env file
run src/main.py
verify result in data folder or in terminal or in index.
Note : If time permits, I am going enhance further with actual tool and user interface as application.
Technical Highlights
This project implements an Agentic (RAG) system for intelligent customer complaint escalation. It combines structured data reviews.csv with embedding-based semantic retrieval and aggregated with orders.csv perform historical pattern analysis and detect recurring issues. By integrating Azure OpenAI, Azure AI Search, and an autonomous reasoning agent, the system moves beyond traditional semantic search to enable aggregation-driven decision workflows. The result is a scalable, memory-aware escalation framework capable of identifying systemic product issues rather than isolated complaints.
Challenges & Learnings
Honest try on getting deterministic information from user review comments.
Being in organization more than 15 years and worked in web application development (15 years) , data platform engineer (1year)- managing data pipelines and exploring AI (2 years) . What I realised is that a document/long text can be embedded in vectorDB to get semantic search, but it can't be taken for data aggregation or as a data point. RAG is probabilistic. Aggregation is deterministic. Trying to find out deterministic information on product and/or category from user review comments.
- Exponential learning curve on azure foundry azure AI search, Azure open AI
- Understanding Azure AI projects, azure open ai and azure-framework sdk was confusing and challenging.
- Which SDK to use when. That decision took significant time to start the project.
Contact Information
https://www.linkedin.com/in/kavita-herur-2207b654/
Country/Region
India