AI-Powered Customer Master Data Management (MDM)
Hey everyone! 👋
I am excited to share that I have officially completed my capstone project as part of the IIT Patna Generative AI program! 🎉
I tackled a real-world challenge: duplicate and inconsistent customer data across multiple systems. My solution is an end-to-end AI-powered Customer Master Data Management (MDM) platform that:
🔹 Detects duplicate and partial-match records using ML embeddings
🔹 Generates a Golden Customer Record
🔹 Enables fast and accurate semantic search
✨ Key Features:
- API data ingestion
- Scoring and confidence interval computation using Drools rule engine
- Embedding-based similarity matching using Sentence Transformers
- k-NN semantic search with OpenSearch
- Data Stewardship View
🛠️ Tech Stack:
- replit.com + Java Spring Boot (Backend)
- OpenSearch for search + k-NN
- Sentence Transformers (all-MiniLM-L6-v2) for embeddings
- lovable.dev (Front End)
- Drools Rule Engine
🎯 Outcome:
✔ Reduced duplicate customer entries
✔ Better customer identity resolution
✔ Faster search (<1 sec) with semantic understanding
This project gave me hands-on experience solving a real enterprise data challenge, and I am proud of the results! 😄
Happy to share demo / architecture if anyone is interested! 🚀
Top comments (0)