Our mission
We build and open source AI tools that advance small models, harness engineering, evolving agents, responsible AI, MLOps and graph machine learning for the financial services industry. By contributing back to the open source ecosystem we help raise the bar for trustworthy AI in banking — and we give back to the community whose work powers our own innovation.
Featured projects
| Project | Description | License | Status |
|---|---|---|---|
ralph |
A configurable Bash/PowerShell loop that runs an AI coding CLI with a fresh session each iteration. | Apache-2.0 | ✅ Active |
ralph-vault-skill |
Skill to generate the knowledge vault for projects using the Ralph loop. | Apache-2.0 | ✅ Active |
auto-bayesian |
Config-driven, interpretable Bayesian network training for relational tabular data. | Apache-2.0 | ✅ Active |
autoguardrails |
Alignment-research scaffold (autoresearch-style) for LLM guardrails over a single policy.md surface. |
Apache-2.0 | ✅ Active |
causal-perception-implementation |
ML research code for causal perception — comparing competing structural causal models via interventional and counterfactual distributions, applied to fair credit decisions. | Apache-2.0 | ✅ Active |
gen-fraud-graph |
Synthetic fraud graph generator for training and benchmarking graph-based fraud detection models. Scales to 100M+ accounts. | Apache-2.0 | ✅ Active |
genetic-algorithm |
A dependency-free Python genetic-algorithm engine with pluggable fitness criteria — a reusable search core for an LLM/AI autoresearcher. | Apache-2.0 | ✅ Active |
linear-adapter-trainer |
Train linear embedding adapters with triplet loss to align retrieval embeddings with your queries (RAG). | Apache-2.0 | ✅ Active |
llm_bridge |
A tiny, vendor-neutral LLM client library — one interface with pluggable adapters for OpenAI, AWS Bedrock and Google Gemini, or bring your own backend. | Apache-2.0 | ✅ Active |
mech-gov-framework |
Mechanical Governance for LLM Decisions — model-agnostic governance regimes, hard gates and governance metrics for high-stakes LLM decision systems. | Apache-2.0 | ✅ Active |
mutatis-mutandis |
Situation testing for discrimination analysis with counterfactual comparators — research code for the paper 'Mutatis Mutandis: Revisiting the Comparator in Discrimination Testing'. | Apache-2.0 | ✅ Active |
sota-stressed-datasets |
Open benchmark datasets republished in stressed form to evaluate ML/LLM robustness. Curated by Santander AI Lab. | CC BY 4.0 + Apache-2.0 | ✅ Active |
All projects use synthetic or anonymised data only. No real customer information is published.
Open source governance
Our Open Source Programme Office (OSPO) runs a transparent two-track review for every project considered for public release:
- Fast Track — forks, generic tools, tutorials, datasets, SDKs without business logic. Reviewed by OSPO Lead with automated scans (SLA < 4 hours).
- Full Track — AI models, frameworks with IP, code that touched internal data. Reviewed by a FOSS Review Board (OSPO Lead + Legal + CISO + Architect). SLA 2-4 weeks.
Full policy: GOVERNANCE.md
Contributing
We welcome contributions from everyone. Please read:
CONTRIBUTING.md— how to submit issues and pull requestsCODE_OF_CONDUCT.md— Contributor Covenant v2.1SECURITY.md— responsible disclosure- All contributors agree to our Contributor License Agreement (CLA) on first PR
Contact
- General open source / partnerships: opensource@gruposantander.com
- Security: security@santander.com
- Website: santander.com
- Careers in AI: santander.com/en/careers
Built with ❤️ by AI Labs · Banco Santander · Madrid 🇪🇸
Open code · Responsible AI · For the community