LFSD (Life Financial Services Decisions) is the reference implementation of an open, community-governed framework for responsible, AI-assisted financial decision-support — built so that institutions without large AI teams (community banks, credit unions, CDFIs, and mission-driven fintechs) can adopt inclusive financial technology with governance built in, not bolted on.
This is an early-stage, working proof of concept under active development. It is
open-source (MIT), openly governed, and not a commercial product: no single
entity controls the codebase or roadmap. See GOVERNANCE/.
1. The framework modules (the reusable core)
packages/responsible-ai— a standalone, pip-installable governance layer: consent management and PII redaction, designed to sit in front of any model call. This is the first framework module extracted for independent reuse.- Decision engine — a policy-first, AI-assisted architecture: deterministic
rules and institutional policy constraints govern outcomes; AI assists with
intake, classification, explanation, and decision support. High-impact
decisions are never delegated to a model. See
docs/DECISION_ENGINE.md. - Scoring dimensions — transparent, explainable scoring across three dimensions of financial wellbeing: time, wealth, and health — the basis for personalized, auditable guidance rather than black-box recommendations.
2. Helmory (the demonstration deployment) Helmory is the consumer-facing demo that exercises the framework end-to-end — onboarding, financial goals, and day-to-day decision support. It exists to prove the architecture works in operation; the framework, not the demo, is the contribution.
- Policy-first, AI-assisted. Deterministic controls above probabilistic models. The system never fabricates data and never auto-executes high-impact financial decisions.
- Governance as architecture. Consent, auditability, explainability, and human oversight are framework components, not compliance afterthoughts.
- Built for resource-constrained institutions. Modules are independently adoptable, vendor-neutral, and designed for institutions that consume technology through vendors rather than building it.
- Open by default. MIT-licensed, community-governed, with security
scanning on every PR (
SECURITY.md).
Current stage: research-and-maturation. The inclusion-focused modules (onboarding, financial goals, financial-wellness scoring) lead; a credit-relevant decision-support module is a research objective requiring fair-lending controls, model-governance review, and regulator engagement before any deployment.
Contributions, issues, and adoption experiments are welcome.