Laboratory · Vol. 01 · 2026
Project Index
Six open-source, production-grade AI systems for fintech data teams, across GenAI, Risk, Causal Inference, Payments, and Experimentation. Each project ships with the business problem, an architecture walk-through, and the working code on GitHub. Select a project to open it.
[GENAI-001]
GenAI
→
BSA/AML Regulatory RAG
Retrieval-augmented generation for querying BSA/AML regulatory documents. Plain-English answers cited to section and page, with hybrid retrieval and deterministic output.
LangChain · Gemini 2.0 · ChromaDB · BM25
[RISK-001]
Risk
→
Customer Churn with Explainability
Production-grade churn prediction for banking customers with automated feature engineering, Bayesian optimization, and SHAP-based explainability.
XGBoost · Optuna · SHAP
[CAUSAL-001]
Causal Inference
→
Credit Limit Intervention Analysis
Causal inference estimating the true effect of credit limit changes on default risk, with refutation tests and heterogeneous treatment effects.
DoWhy · EconML · Causal Forest
[FRAUD-001]
Payments
→
Imbalanced Classification Benchmark
Nine techniques for extreme class imbalance in fraud detection, compared on the same split with PR-AUC: resampling, class weighting, anomaly detection, threshold tuning.
XGBoost · imbalanced-learn · SHAP
[EXPD-001]
Experimentation
→
A/B Test Design Pipeline
Turns a raw customer base into two balanced A/B test groups: power analysis for sample size, stratified randomization, and statistical balance validation.
statsmodels · SciPy · scikit-learn
[MCP-001]
GenAI
→
Fintech Data MCP Server
Template architecture for an MCP server that gives any LLM structured, read-only access to an internal database through predefined parameterized queries.
MCP · SQLAlchemy · Pydantic