Large-scale manufacturing
Industrial analytics platform
An AI and data platform that turned factory sensor streams into actionable predictions—designed, built, and deployed as a production MVP for one of the largest manufacturers in its sector.
- ML pipeline
- Cloud
- Industrial
- Data engineering
Web UI
Backend API
AI layer
Data store
Context
A ceramics manufacturer needed more than dashboards—they needed models that could learn from live production data and support decisions on the factory floor. Sensor data lived in silos, experiments were hard to reproduce, and there was no clear path from prototype to something operators could rely on daily.
What we built
We designed a unified data layer, ETL pipelines from hundreds of factory sensors and external APIs, and a training workflow to iterate on many model candidates systematically. The MVP connected ingestion, model training, evaluation, and cloud-hosted inference—with outputs surfaced to the factory management tools leaders already used.
Delivery highlights
- Unified 300+ sensor variables into a managed database with documented schema
- Built Python and SQL ETL pipelines feeding dashboards and operational exports
- Trained and evaluated 10+ machine learning models with reproducible experiment tracking
- Deployed selected models for real-time inference on cloud virtual machines
- Documented all pipelines and code for handoff and ongoing iteration
Impact
- MVP adopted by factory leadership as a foundation for operational AI decisions
- Reduced friction between data collection, model experimentation, and deployment
- Established a scalable pattern for adding new sensors and model types over time