Disclaimer: Independent educational project. Not affiliated with JPMorgan Chase. Built by Kaushal Khodifad. Data from public sources; figures are estimates.return to portfolio
Disclaimer: Independent educational project. Not affiliated with JPMorgan Chase.

Reference architecture

The platform that makes it possible

How a regulated card org turns raw network files into governed, entitled data products - and runs ML and GenAI on top without ever loosening control. Built on a data-product mesh; the live Postgres RLS plane at the bottom is the one runnable proof.

2.1

The spine - four paradigms, layered

Each layer answers a question a card org has to answer for a regulator. Together they are the mesh.

Raw network files
auths · settlement · disputes
Bronze → Silver → Gold
medallion refinement
Data products + contracts
domain-owned · ODCS
ML pipelines
feature tables · model cards
GenAI on curated data
NL-to-SQL · governed RAG

The entitlements/RLS stage is the one runnable proof - click it to open the live simulator.

L1
Lakehouse

One ACID store for raw auth streams and curated products - no warehouse/lake split to reconcile at examiner time.

Delta Lake / open table formatObject storage (S3)Spark / Photon computeStreaming ingest (Auto Loader)
L2
Medallion (Bronze / Silver / Gold)

A disciplined refinement path from raw network files to certified products - replayable, quarantined, contract-validated.

Bronze append-only landingSilver conformed + validatedGold certified products & feature tablesDeclarative pipeline expectations
L3
Data mesh / data products

Card domains (spend, rewards, risk, partner economics) own and publish their own products; no central team bottleneck.

Domain ownershipData-as-a-productSelf-serve platformFederated computational governance
L4
Federated governance + entitlements

Policy is computed and enforced uniformly - row scope, column masking, purpose, lineage, audit - so a regulated bank can prove control.

ABAC row/column policiesGoverned tags + classificationAutomatic lineageSystem/audit tables for examiners
01
Domain ownership

Decentralized ownership of analytical data by the domain that knows it best.

02
Data as a product

Each domain serves its data as a discoverable, addressable, trustworthy product.

03
Self-serve data platform

A platform that lets domains build and run products without bespoke plumbing.

04
Federated computational governance

Global policies (security, interop, quality) enforced as code across all products.

Data-mesh four principles - Dehghani / Thoughtworks canonical formulation.

2.2

Medallion, mapped to card data

The Gold tables ARE the published data products - the same ones in the live catalog.

Bronze
Bronze - raw, replayable

Network and ledger events land exactly as received. PAN tokenized at landing.

  • Card auths + settlement/clearing (ISO 8583 / ISO 20022)
  • Visa / Mastercard network files
  • Disputes & chargebacks, rewards events
  • PAN tokenized at landing; append-only; replayable
  • Ingested via streaming / Auto Loader
Silver
Silver - conformed, validated

Cleaned, deduped, joined to reference data, and validated; bad rows quarantined.

  • Joined to merchant / MCC / account / cardholder reference
  • Deduped & type-conformed
  • Validated with declarative pipeline Expectations (DLT / Spark Declarative Pipelines)
  • Non-conforming rows quarantined, not silently dropped
Gold
Gold - the published products

Certified aggregates and ML feature tables. These Gold tables ARE the data products in the catalog.

  • Certified business aggregates (spend, approval, fraud, attrition)
  • ML feature tables with online/offline parity
  • Contract + SLA + entitlement attached per product
  • Consumed by BI, ML, partner reporting, and GenAI - never the raw lake
See the Gold products in the catalog
2.3

Domain-owned data products + data contracts

A data product = data + ownership + SLA + contract + consumers. A product without a contract is just a table.

Trusted-data-product characteristics
DiscoverableAddressableTrustworthySelf-describingInteroperableSecureQuality & observabilitySemantic consistencyPrivacyDiscoverability
Lifecycle
InceptionDesignCreationPublishOperate / GovernConsumeRetirement
Open Data Contract Standard (ODCS) v3.1

Bitol project, Linux Foundation AI & Data - originated at PayPal

FundamentalsSchemaData-quality rulesSLAServersTeamAccess control

Open Data Product Standard (ODPS) v1.0 - the product wrapper around the contract

This is what a contract looks like, enforced
2.4

Governance & entitlements - the Chase-scrutinized part

Two enterprise stacks, because a reviewer should see you know both.

DatabricksUnity Catalog
  • ABAC policies

    Row-filter and column-mask policies bound to governed tags - attribute-based, not per-table hand-coding.

  • Governed tags + classification

    Auto-tag PII / financial data, aligned to GDPR / PCI / HIPAA controls.

  • Row filters & column masks

    Enforced uniformly across SQL, ML, and BI consumers.

  • Automatic lineage

    Table- and column-level lineage captured automatically for impact analysis and audit.

  • System tables

    system.access.audit, lineage, and billing tables - examiner evidence by query.

SnowflakeHorizon
  • Masking + external tokenization

    Dynamic data masking and external tokenization integrations at column level.

  • Row access policies

    Row-level filtering bound to roles / context.

  • Tag-based masking

    Object tags drive masking policies so classification propagates automatically.

  • Trust Center

    Posture monitoring across the account.

  • Classification / dependencies / access history

    Data Classification, Object Dependencies, Access History, Data Metric Functions.

PII tokenization · PCI-DSS

Vaultless tokenization + format-preserving encryption (FPE) applied to PAN at Bronze landing. Detokenization is itself an entitlement - granted only to roles and purposes that need clear values. Representative vendor: Protegrity (representative vaultless-tokenization vendor).

Everything above is a slide until you can prove it.

Atlas enforces row scope + purpose + column masking + audit on a real Postgres RLS plane. Open the live simulator - the one runnable governance component.

Open the live simulator
2.5

The JPMC home-turf reference (AWS-native mesh)

JPMorgan has publicly described an AWS-native data mesh. Atlas's design is consistent with it; the firm-level mesh, JADE, and the move to Databricks lakehouse + Snowflake Cortex are the closest public proxy for an internal card stack the firm does not publish.

Physically-isolated S3 lakes

One per data product / producer - hard isolation, not shared buckets.

AWS Glue Data Catalog + mesh catalog

Per-product Glue catalogs roll into a master catalog that tracks every share.

AWS Lake Formation - share in place

Data shared in place (not copied) at column / record / individual-value granularity.

Athena for consumption

Consumers query entitled shares directly; no extract proliferation.

Puddles → ponds → master LOB lake

Producer/consumer accounts roll up to a master line-of-business Lake Formation account.

Office of the CDO + master Glue catalog

Federation and policy enforced centrally; JADE = JPMorgan Advanced Data Ecosystem.

Mapped to card products:Card-AuthDisputesInterchangeRewards
2.6

ML pipelines + GenAI on curated data

GenAI consumes only Gold / certified products through the catalog - never the raw lakes.

Governed RAG
Delta knowledge baseMosaic AI Vector SearchModel ServingMLflow + Unity Catalog governance
Text-to-SQL / NL analytics
Databricks AI/BI GenieGenie OntologyUC Metric Views (governed metric definitions)Snowflake Cortex Analyst (equivalent)

Atlas's live /console is exactly this: a Genie-style governed text-to-SQL surface running under the same RLS plane.

Guardrails
Mosaic AI GatewayPII redactionRate limiting + safety filtersModel allowlistTelemetry: cost / latency / guardrail-trigger rate

Mirrored in the console's model allowlist + SELECT-only run_select guard - the same idea, enforced in this demo.

See the live Genie-style console
2.7

Model risk - SR 11-7 → 2026 guidance

On April 17, 2026 the Federal Reserve, FDIC, and OCC replaced SR 11-7 (and OCC 2011-12 / FIL-22-2017) with a risk-based, principles-driven model-risk framework. GenAI and agentic AI are formally OUT OF SCOPE of SR 26-2 (the agencies deemed them novel and rapidly evolving, with an RFI to follow) - but they inherit the same MRM principles by analogy.

ShiftOld - SR 11-7 (2011)New - SR 26-2 (Apr 2026)
ScopeOne-size-fits-all validation for all modelsRisk-based tiering - controls scale with materiality
CadenceValidation as a point-in-time gateLifecycle thinking - risk managed end-to-end
ChallengeIndependent validation functionEffective challenge remains the backbone
MonitoringPeriodic revalidationContinuous drift & outcome monitoring in prod
AI / GenAINot addressedOut of scope of SR 26-2 - inherits MRM principles by analogy; RFI to follow
01
Risk-based tiering

Controls scale with model materiality and risk - not one-size-fits-all.

02
Lifecycle thinking

Risk managed across the full model lifecycle, not at a single validation gate.

03
Effective challenge

Independent, competent challenge remains the backbone of validation.

04
Continuous monitoring

Ongoing performance, drift, and outcome monitoring in production.

05
AI inheritance (by analogy)

GenAI / agentic systems are out of scope of SR 26-2, but inherit the same MRM principles by analogy until dedicated AI guidance follows.

RequirementComponent
Model inventoryUnity Catalog tables / model registry
Model cardsMLflow + UC Model Registry, bound to the production version
Validation / effective challengeMLflow Evaluate: AUC / KS / PSI; LLM: groundedness, toxicity, PII-leak, LLM-as-judge
Continuous monitoringLakehouse Monitoring + Gateway telemetry
Promotion controlsUC Registry aliases + ABAC; Tier-1 dual control
Examiner RFIAnswered via a governed Genie space over the inventory + validation log
§9

Component cheat-sheet

The full name-drop set, grouped by layer.

Storage / compute
Delta LakeS3Spark / PhotonAuto LoaderLakehouse
Mesh fabric
Data meshDomain productsGlue Data CatalogJADE
Governance / entitlements
Unity Catalog ABACSnowflake HorizonLake FormationImmuta (optional)Protegrity FPE
Contracts / products
ODCS v3.1ODPS v1.0Metric Views
Sharing
Delta SharingLakehouse FederationClean RoomsMarketplace
GenAI
Mosaic AI Vector SearchModel ServingAI/BI GenieMosaic AI GatewayCortex Analyst
Model risk
MLflow EvaluateUC Model RegistryLakehouse MonitoringPR-AUC / KS / PSI

Sources & methodology

Real · live Public-sourced · cited Synthetic · illustrative
  • 01
  • 02
  • 03
  • 04
  • 05
  • 06
  • 07
Synthetic dataFictional partnerIllustrative metricsGrounded-inference topology

↑ Independent educational project by Kaushal Khodifad. JPMorgan Chase is a real company in the card data & analytics space; this design and the underlying prototype were built by Kaushal as a portfolio study. Not affiliated with, endorsed by, or representative of JPMorgan Chase's actual product. Data is from public sources. Figures are estimates.

Return to portfolioOpen the live demo