IBM Watsonx Gen AI Challenge Project – Service Case

This repository contains a sanitized version of the solution submitted for the IBM Watsonx Gen AI Challenge (2024).
The project addresses a real-world use case provided by Komax, a global market and technology leader in automatic and semiautomatic wire and cable processing solutions.
All data shared here is synthetic and intended for demonstration purposes only.

⚠️ Disclaimer: Client-provided email threads and know-how article examples are strictly confidential and not included in this repository.
The examples here are fabricated to reflect the structure and content of actual service interactions.

---

Business Problem

Scenario
The aim is to transform past service emails into structured know-how articles. These emails typically consist of multi-turn dialogues between Komax service technicians and customers, often in German or English.

Users
Internal technical service team at Komax

Pain Points

• Large volume of unstructured email data
• Manual review is infeasible
• Complexity exceeds the capabilities of traditional rule-based solutions

---

Solution

Goal
Use large language models (LLMs) to automatically generate high-quality know-how articles from service email exchanges.

Scope
The LLM must:

• Understand multi-turn German and English email threads
• Extract relevant case data (e.g., issue, resolution, affected components)
• Fill a predefined structured output format (e.g., Excel)

---

Project Overview

Input
Unstructured multi-turn email threads
📩 Example: A customer reports a machine error, and the technician replies with troubleshooting steps and a follow-up request.

Output
Structured table with fields such as:

• Case Number
• Owner
• Created/Modified Dates
• Error Description
• Root Cause
• Solution
• Status

---

Core Features

• 🔄 Few-shot prompting with realistic synthetic examples
• 🤖 Prompt-based extraction using IBM Watsonx LLMs
• 📄 Output formatted as Excel/CSV for downstream use
• 🌍 Multilingual support (German and English)

---

📁 Files

Path	Description
notebooks/	Cleaned Jupyter notebooks for processing email threads
data/	Synthetic input email samples
output/sample_output.xlsx	Example structured output file

---

Run Instructions

1. Insert your IBM Cloud project credentials into the notebook.
2. Run the notebook to call the Watsonx LLM API.
3. Output will be saved as a CSV or Excel file in Watson Studio.

---

License

This repository is intended solely for demonstration and educational purposes.
No client or proprietary data is included.