MDMP Specification (Draft)¶

The IINTS MDMP is a data-governance layer for AI research pipelines.
It standardizes how datasets are validated, fingerprinted, and scored before model training or evaluation.

For the complete operational guide (commands, rules, scoring, gates, and pitfalls), see MDMP_FULL_GUIDE.md.

MDMP is designed for traceability and reproducibility in research settings.
It does not certify a model for clinical use on its own.

MDMP has a dedicated SDK namespace: - Python: iints.mdmp - CLI: iints mdmp ...

Who This Page Is For¶

Researchers defining pre-model data quality gates.
Engineers implementing reproducible data pipelines.
Reviewers checking traceability and dataset provenance evidence.

Terminology Used Consistently In This Page¶

Contract: machine-readable schema + constraint definition.
Validation run: one execution of contract checks against a dataset.
Grade: MDMP level (draft, research_grade, clinical_grade).
Fingerprint: deterministic SHA-256 hash for contract and dataset.

Section Structure¶

This page is structured as: - Purpose - When to use - Commands - Output / Artifacts

Scope¶

Purpose - Define what MDMP currently governs inside IINTS.

When to use - Before model training, fine-tuning, benchmarking, or report generation.

MDMP currently covers: - tabular/time-series ingestion via data contracts - structural and range validation - deterministic dataset fingerprinting - reproducible compliance scoring

Implemented in: - iints data contract-template - iints data contract-run

Core Pillars¶

Structural Integrity
Contracts define required columns, expected types, and physiologic ranges.
Unit conversions can be applied deterministically (for example mmol/L -> mg/dL).
Chain of Trust
Contract and dataset fingerprints are generated with SHA-256.
Any source-data change produces a new fingerprint.
Compliance Scoring
Every run returns compliance_score and pass/fail checks.
MDMP grade is derived from score + hard compliance.
Runtime Guardians
mdmp_gate can enforce MDMP checks directly before function execution.
This supports in-memory pipeline protection for research workloads.

MDMP Grades¶

clinical_grade: compliant and score >= 90
research_grade: score >= 75
draft: below research threshold

Grade gating can be enforced in CI:

Commands

iints data contract-run data_contract.yaml data/my_cgm.csv \
  --min-mdmp-grade research_grade \
  --fail-on-noncompliant

Certification dashboard generation:

iints data mdmp-visualizer results/contract_data_report.json \
  --output-html results/mdmp_dashboard.html

Synthetic mirror generation:

iints data synthetic-mirror data/real.csv contracts/clinical_mdmp_contract.yaml \
  --output-csv data/synthetic_mirror.csv \
  --output-json audit/synthetic_mirror_report.json

Runtime function gate:

from iints import mdmp_gate

@mdmp_gate("contracts/clinical_mdmp_contract.yaml", min_grade="clinical_grade")
def process_dataframe(df):
    ...

Recommended Workflow¶

Output / Artifacts - contract_data_report.json / mdmp_report.json - MDMP grade + compliance score - contract and dataset fingerprints - optional HTML dashboard for audit sharing

Write data_contract.yaml for your dataset.
Run iints data contract-run and store JSON output.
Archive fingerprints with model artifacts.
Train/evaluate only on datasets meeting your minimum grade policy.

Audit Artifacts¶

Keep these together per run: - contract_data_report.json - validation_report.json - sources_manifest.json - SUMMARY.md - model artifact metadata (including seed, split, and commit SHA)

Regulatory Positioning¶

MDMP is a research data-quality protocol that can support documentation practices expected by regulated environments.
It is not itself a regulatory approval and does not turn the SDK into a medical device.