PRISM architecture, purpose-built for financial services
Progressive data refinement from Protected Raw ingestion through Refined transforms and Intelligence enrichment to Market-Ready outputs. Quality gates at every layer transition.
P → R → I → S → M
Five layers of progressive data refinement. Quality gates enforce standards at every transition — only data that passes is promoted to the next layer.
Protected Raw
Compliance-first ingestion with PII scanning and immutable audit trail.
- Compliance-first ingestion — PII scanned before storage
- Immutable audit trail for every record
- Format auto-detection on ingestion
- Append-only storage with full history
{ "customer_id": "C-4521", "name": "Acme Corp", "txn_date": "2024-01-15", "amount": "12450.00", "status": null}Structured, semi-structured, and unstructured
Handle every data format your enterprise produces. From relational tables to PDFs, JSON feeds to server logs.
Structured
Full schema discovery and type mapping
Semi-Structured
Schema inference and nested field flattening
Unstructured
OCR/NLP text extraction and metadata extraction
Format Auto-Detection
Automatically identifies file formats on ingestion
Cross-Format Migration
Convert between relational, Parquet, JSON, and more
Schema Inference
Infer schemas from semi-structured and unstructured data
Pre-built for financial services
Start with a battle-tested template and customise to your needs. Each template ships with pre-configured PRISM layers.
Financial Transactions
KYC/AML pipeline with transaction monitoring and sanctions matching
Protected Raw
Raw transaction feeds, KYC documents, sanctions lists
Refined Transforms
- Deduplication
- Type casting
- Null handling
- Schema validation
Market-Ready Outputs
- Transaction risk scores
- Sanctions match flags
- Suspicious activity reports
Customer Master
Customer 360 with MDM, ID resolution, and lifecycle tracking
Protected Raw
Raw CRM exports, support tickets, usage logs
Refined Transforms
- Identity resolution
- Address standardization
- Dedup/merge
- SCD Type 2
Market-Ready Outputs
- Customer 360 profiles
- Lifetime value scores
- Churn predictions
Generic
General-purpose PRISM template with SCD Type 2 and time-series support
Protected Raw
Any structured, semi-structured, or unstructured source
Refined Transforms
- Schema validation
- Type casting
- Deduplication
- Null handling
Market-Ready Outputs
- Star schema
- Aggregate tables
- Time-series rollups
From visual canvas to production workflow
The pipeline compiler transforms your visual DAG into a durable Temporal workflow in five deterministic steps.
Validate DAG
Check for cycles, verify all required inputs are connected, validate component compatibility
Resolve Components
Lock component versions, resolve configuration schemas, validate parameter types
Apply Policy Overlays
Merge organisation → team → pipeline security and compliance policies
Generate Workflow
Compile visual DAG into Temporal workflow definition (Python)
Register & Emit Lineage
Register with Temporal server, emit column-level lineage metadata as byproduct
Compilation Path
Visual DAG → Intermediate Representation → Temporal Workflow
Column-level lineage metadata is emitted as a byproduct of compilation, not a separate process.
Start with a template, customise with the visual builder
Choose from pre-built financial services templates or build from scratch. Your pipeline, your way.