Industrial Asset Management & Analytics Platform

The Problem

ICAR needed a scalable way to manage hierarchical industrial asset data, ingest and process large volumes of time-series measures, compute derived metrics across multiple aggregation levels, and expose analytics quickly enough for operational and reporting use cases. The challenge was to combine high-performance computation, real-time ingestion, flexible referential modeling, and controlled infrastructure costs.

What We Built

• Serverless industrial data platform combining referential asset data, raw measures, computed measures, and analytics views
• Hierarchical asset and plant modeling in Firestore with support for historical changes, dynamic attributes, and portfolio grouping
• Time-series storage architecture in Bigtable optimized for real-time reads and writes at industrial scale
• In-memory computation layer using Pandas, NumPy, and NetworkX for high-performance matrix-based calculations and metric dependency graphs
• Dynamic formula engine able to define and compute metrics across levels using attributes, measures, aggregation rules, and calculation dependencies
• Apache Beam streaming pipelines consuming Pub/Sub messages, processing raw files, aligning sampling, computing measures, and writing results back to Bigtable
• BigQuery analytics layer with external and internal tables, views, and materialized views for faster reporting access
• Visualization app built with Plotly and Streamlit, deployed on Cloud Run
• Locally testable environment with mocked services, automated GitHub Actions pipelines, and 93% unit test coverage

Tech Stack