Real-time data warehouse & ML platform
AIT · Technology / Data Infrastructure
Latency benchmarked pre/post migration; accuracy measured against the legacy reporting baseline.
- Client
- AIT
- Industry
- Technology / Data Infrastructure
- Scope
- Legacy-to-cloud data-platform migration with real-time streaming and custom ML.
Inside the project
Challenge
AIT needed a scalable, real-time analytics infrastructure to replace their legacy SQL Server and support ML-based decision-making.
Solution
We migrated AIT's databases from SQL Server to Google BigQuery, using Kafka Connect for real-time streaming with change-data-capture, deployed a Kubernetes cluster for scalable processing, and built custom ML models on Google Vertex AI.
Our approach
We migrated AIT from SQL Server to Google BigQuery, streamed changes in real time with Kafka Connect (change-data-capture), ran scalable processing on Kubernetes, and trained bespoke models on Vertex AI. The rollout followed a governed sequence — inventory SQL Server workloads, prioritize manufacturing use cases and map schemas, establish Kafka CDC streams into validated BigQuery datasets, deploy Kubernetes services, build and test Vertex AI models, then add monitoring, access controls and operational support — keeping ingestion, curated data, model services and business consumption as separate permission layers with full lineage from source tables through dashboards to predictions. So machine, production, service and commercial data operate as one governed decision system, not disconnected reports.