Your historian holds a decade of operational truth. The constraint has never been collection — it is that the archive cannot be joined to maintenance records, lab results, weather or market prices at a scale where statistical methods start to pay.

That constraint has come apart. AVEVA CONNECT now integrates with Databricks over Delta Sharing, and the Lakeflow stack has reduced what used to be a custom connector programme to configuration and data modelling. LAMR builds the modelling.

What we deliver

  • Architecture and ingestion design. Delta Sharing, Lakeflow Connect, or streaming — selected against your estate, your network boundary and your latency requirement rather than a default pattern.
  • Medallion pipelines tuned for time series. Bronze landing raw tag history replayable and at full fidelity; silver carrying asset context, normalised engineering units and preserved quality codes; gold holding the engineering features the business actually asks for.
  • Context propagation. Carrying the PI Asset Framework hierarchy into Unity Catalog, so analysts query assets rather than tag names. This is the step that separates a data product from an expensive archive.
  • Historian-aware handling. Watermarking, backfill and replay for late and out-of-order data; interpolation of exception-based samples onto even intervals so windowed aggregation means something.
  • Governance. Unity Catalog access control, lineage from a gold KPI back to its source tag, and audit trails that hold up for regulated reporting.
  • Performance. Clustering on asset and timestamp, so a five-year trend returns in seconds instead of scanning the table.

Why this needs both skill sets

This work sits at an awkward intersection. It requires people who understand Asset Framework hierarchies, exception-based compression and why a deadband exists — and the same people need to be fluent in Delta Lake, pipeline design and Unity Catalog governance. OT engineers do not usually write Spark. Data engineers do not usually know what a digital state is.

Projects that fail generally fail the same way: tags lifted without the asset model, OT data resampled as though it were IT data, governance retrofitted, and no path back to the control room for the insight once it exists.

Where it leads

The lakehouse is a means, not the point. What it enables is condition-based maintenance across a fleet rather than one unit, degradation models trained on full history, and operational and financial reporting that finally agree on one number.