Lakehouse Platform + AI Speed Layer Part of the Governed, Unified, Model-Driven Real-Time Data Platform Suite (GUM-RTDP) — Source-code licensed reference architecture.

What It Does

The Lakehouse Platform transforms raw business events into trusted, governed analytical assets — in near real time. Built on a medallion architecture (Bronze → Silver → Gold), it delivers clean dimensional data to dashboards and APIs with the response times your business demands.

The AI Speed Layer is a distinct, independently licensable platform that scores business events in real time, enabling real-time AI-driven decisioning across a broad range of use cases. Fraud detection and anomaly identification are included as reference examples. It consumes business events directly from the event fabric — the same events produced by the EDA Platform — operating independently of, and alongside, the Lakehouse Platform.

Medallion Architecture

  • Bronze Zone — raw event ingestion from Kafka topics into MinIO object storage via incremental micro-batch processing. Immutable, auditable, replayable.
  • Silver Zone — business entity integration layer built on the CEDM (Conceptual Enterprise Data Model). Raw events are transformed into clean, governed entities — customer, product, order — with temporal tracking preserving full history.
  • Gold Zone — dimensional consumption layer built on the DEDM (Dimensional Enterprise Data Model). Snowflake schema (or denormalized snowflake) ready for BI dashboards and analytical APIs with sub-second response times via an in-process serving layer.

Serving Layer

  • The Gold Zone (and sometimes the silver zone) is partly replicated in the serving layer which is essentially a low latency DB built for speed which is used by dashboards and API. The reference architecture is using an in-process analytical engine that reads Delta tables directly from object storage and creates a replicated DB with views delivering dashboard and API response times. This architecture is tool-agnostic — it does not impose BI or low latency DB such as Snowflake, Synapse, Tableau, or Power BI, preserving your freedom of choice.

AI Speed Layer

The AI Speed Layer is a real-time intelligence engine that scores business events as they occur, consuming directly from the event fabric. The reference implementation demonstrates two example models:

  • Fraud Detection — identify fraudulent order patterns, in real-time, before they are processed, not hours later
  • Anomaly Detection — flag statistically abnormal transactions that fall outside expected business behavior in real time

These are reference examples contextual to an order system — many other AI models can be implemented following the same pattern. See the Industry Use Cases for details.

Inference is performed using ONNX-based models (Random Forest for fraud scoring, Isolation Forest for anomaly detection) running in a dedicated Go-based inference engine. Results are published back to the event fabric as first-class business events — consumed by the Order Management application for threshold evaluation and human review escalation.

End-to-end latency: under 1.5 seconds (as low as 0.8 seconds in Direct EDA Mode) from event to threshold processing of scored result.

For context — as a customer of a major Canadian bank, a potentially fraudulent transaction on one of our accounts was flagged 24 hours after it occurred. Your organization can do it in under 1.5 seconds. — detecting and blocking fraud before it completes, sparing customers the burden of disputes, chargebacks, and compromised accounts. Real-time detection is no longer a competitive advantage — it is a customer expectation.

What the Lakehouse Platform Includes
A full reference implementation — incremental PySpark scripts, compensation logic in the gold zone (an accounting-style process that reverses prior records through new entries, preserving full historical integrity while keeping current metrics accurate), container stack, scheduling, OLAP dashboard, Visual Paradigm models, Kafka topics, Delta Lake tables, utilities, demo assets, and documentation covering architectural aspects.

What the AI Speed Layer Includes
The ONNX-based inference engine (Go), the dedicated process that persists AI events from Kafka into its own AI events store table, and supporting documentation.

Technology Foundation

Apache Kafka · Apache Spark · Delta Lake · MinIO · ONNX · Go · Python · PySpark — each proven individually at scale by Tier 1 technology organizations. The Lakehouse Platform is the only reference architecture that integrates all of them into a governed, real-time lakehouse, complemented by the AI Speed Layer for real-time AI-driven decisioning.

The Lakehouse Platform and the AI Speed Layer are each part of the GUM-RTDP suite — alongside the EDA Platform and Data Governance Platform — and can be acquired independently or together. Source-code licensed for enterprise deployment.

To learn more about pricing, licensing options, and acquisition modalities, contact us at software@itarchitectureandstrategy.com — we look forward to hearing from you.