Fin-XAI is a distributed framework that bridges the gap between high-performance enterprise .NET applications and advanced Python-based Machine Learning.

Machine Learning models in finance are often treated as "Black Boxes." This project provides a novel architecture combining a .NET 8 Blazor Razor Class Library (RCL) with the analytical depth of Python's SHAP (Shapley Additive exPlanations) to provide real-time, transparency-first decision auditing for financial portfolios.

The solution is divided into two primary environments:

1. Inference Engine (Python / FastAPI): A lightweight API that keeps a Random Forest classifier (trained on the UCI German Credit Dataset) loaded in memory. It calculates dynamic SHAP values on the fly to explain why a specific financial decision or risk score was assigned.
2. Client Dashboard (FinXAI.Universe NuGet / Blazor): A plug-and-play Razor Class Library containing the data models, HTTP services, and a fully interactive UI component (<FinXaiDashboard />). It uses JS Interop to handle client-side PDF/CSV exports without taxing the server.

To run the full simulator locally, you need both the Python engine and the .NET host running.

Ensure Python 3.9+ is installed. Navigate to the FinXAI.Universe/PythonEngine directory.

# Install required ML and API libraries
pip install fastapi uvicorn pandas scikit-learn ucimlrepo shap

# Start the API server on port 7001
python Main.py

Open the solution in Visual Studio. Set FinXAI.Demo as the Startup Project and run the application (F5). Navigate to the /finxai route to interact with the dashboard.

This project is packaged as a reusable Blazor UI library. To add the Explainable AI dashboard to your own .NET 8 application:

1. Install the package:

dotnet add package FinXAI.Universe

2. Register the required HTTP Client in your Program.cs:

builder.Services.AddScoped(sp => new HttpClient());

3. Drop the component into any Razor page:

@using FinXAI.Universe.Components

<FinXaiDashboard ApiBaseUrl="[http://127.0.0.1:7001](http://127.0.0.1:7001)" />

• Frontend/UI: .NET 8, Blazor Server, Bootstrap 5, Chart.js (via ChartJs.Blazor)
• Backend/API: Python, FastAPI, Uvicorn
• Data Science: Scikit-Learn, SHAP, Pandas, UCI Machine Learning Repository

Yash Saini