Operating System for AI & Data Governance in Financial Institutions.
A practical framework for building audit-ready, compliant and production-safe data science and machine learning workflows in regulated environments (LGPD, BACEN, CVM aligned).
Financial institutions are rapidly adopting AI and advanced analytics, but most environments still lack:
- clear AI governance workflows
- audit-ready ML pipelines
- controlled data access for training
- reproducible and explainable models
- safe storytelling and reporting practices
This repository provides a structured operating model for governing data, analytics and machine learning in regulated environments.
1. Data never travels unnecessarily
Compute must go to the data. Outputs must be aggregated, controlled and auditable.
2. Governance by design
Compliance and auditability are embedded into the workflow, not added later.
3. Evidence-first pipelines
Every critical step produces structured evidence for audit and traceability.
4. Reproducible and explainable ML
Models must be transparent, documented and defensible.
5. Secure storytelling
Dashboards and reports must never expose sensitive data.
- AI & Data Governance operational framework
- Multi-agent governance structure
- Reusable skills for regulated ML workflows
- Evidence and audit templates
- SQL patterns for governed datasets
- Model validation and approval flows
- Safe BI and storytelling patterns
Designed for:
- banks
- asset managers
- fintechs
- regulated analytics environments
- Data classification and sensitivity mapping
- Legal basis and compliance validation (LGPD/BACEN/CVM)
- Access control and audit logging
- Governed dataset creation (SQL views)
- Data quality and lineage validation
- Protected model training (compute-to-data)
- Model selection and validation
- Evidence generation and model card
- Approval and production deployment
- Monitoring and safe reporting
agents/ → governance and ML agents (10 core + Databricks)
skills/ → operational governance skills (29 finance + 5 Databricks)
workflows/ → reusable governance workflows
rules/ → compliance and governance rules
docs/ → operating manuals, architecture docs
templates/ → evidence, policy and report templates
demos/ → synthetic examples and use cases
install/ → shell installers (local & global)
cli/ → npm CLI installer (npx @fabioforest/fgos-kit init)
docs_site/ → MkDocs documentation site
.agent/ → IDE runtime (Cursor, VSCode, Codex, Antigravity)
git clone https://github.com/fabioffigueiredo/finance-data-governance-os.git
cd finance-data-governance-os
bash install/local/install.shgit clone https://github.com/fabioffigueiredo/finance-data-governance-os.git
cd finance-data-governance-os
bash install/global/install.shnpx @fabioforest/fgos-kit initSee INSTALL.md for full installation details.
- Credit risk modeling with audit-ready pipelines
- Fraud detection governance
- Customer analytics under LGPD constraints
- Model validation and regulatory reporting
- Executive dashboards without sensitive data exposure
Data scientists, ML engineers, data engineers and technical leaders working in regulated environments who need structured and defensible AI workflows.
v1 — Governance OS (current)
v2 — MCP Server + Advanced ML governance modules
v3 — Enterprise governance platform (future direction)
The installer is enterprise-grade safe:
- Dry-run:
npx @fabioforest/fgos-kit init --dry-run(simulate without changes) - Audit Logs: detailed execution logs saved to
.agent/_audit/ - Auto-backup: overwrites automatically create backups (e.g.,
.agent.bak-2024...) - Safe-by-default: incremental updates (adds missing files, preserves existing ones) unless
--overwriteis used
See Safety Policy for details.
Fabio Ferreira Figueiredo
AI, Data & Infrastructure Engineer
Focus: AI governance, data platforms and regulated environments