The challenge
A global construction tools company wanted its sales teams worldwide to manage business opportunities proactively, guided by AI-driven recommendations inside a new application. Account managers needed to know which customers to meet, which opportunities to pursue, where orders stood and how their products were stocked in customer inventories, and they needed AI chatbots that could answer those questions accurately.
That required a data & analytics foundation on Microsoft Fabric that was trusted, confidential and production-ready: a scalable architecture aligned with data federation principles, a semantic layer AI could reason over, and data products with the quality controls to keep every recommendation accurate and relevant.
What we did
Designed and implemented a scalable data architecture in Microsoft Fabric, aligned with data federation principles while managing the non-functional requirements on data quality, security, governance and confidentiality.
- Designed Bronze, Silver & Gold data layers following modular, domain-oriented design principles.
- Defined governance, security and data quality standards to ensure trusted, confidential and production-ready analytical assets.
- Collaborated with business and technical stakeholders to evolve the architecture as new AI-powered use cases emerged.
Designed and implemented an AI-ready semantic layer and its underlying data mart, enabling AI chatbots to give accurate, context-aware answers to the business questions of account managers.
- Modelled business entities, metrics and relationships to provide consistent semantic definitions for analytical questions.
- Designed the underlying data layer to maximize retrieval speed and ensure the quality of the information provided.
- Managed the development of these analytical assets using Claude in an AI-assisted process that delivered more than 10x productivity.
Developed the pipelines, notebooks & data products behind recommending customer meetings, managing opportunities, tracking orders and monitoring the company's product levels in customer inventories.
- Built notebooks in PySpark & Spark SQL and Microsoft Fabric pipelines to ingest, transform and load the data products.
- Worked closely with developers and business stakeholders to enable the downstream analytical capabilities of the application.
- Introduced data quality controls to ensure the accuracy and relevance of the AI-powered recommendations to account managers.