Why should you care?

AIPredict clients need fraud scores and forecasts via JSON, not embedded DLLs.

See it live — copy this example

Use a .NET console or Web API project with Microsoft.ML. Run dotnet run after pasting.

app.MapPost("/api/predict/fraud", (FraudRequest req, PredictionEnginePool<TxRow, FraudPred> pool) =>
{
    if (req.Amount <= 0) return Results.BadRequest("Amount required");
    var p = pool.Predict(new TxRow { Amount = req.Amount, Hour = req.Hour, MerchantCategory = req.Merchant });
    return Results.Ok(new { p.PredictedLabel, p.Probability, model = "fraud-v12" });
});

What happened?

• Validate input DTOs, predict via pool, return label + probability + model version for client debugging.
• Follow the steps below — typing the code yourself is the fastest way to learn.

Practice next

1. Define FraudRequest DTO.
2. Register PredictionEnginePool.
3. MapPost with validation.
4. Add ProblemDetails on errors.
5. Require API key middleware.

Remember

DTO validate. Pool predict. Version in response.

AIPredict fraud API

POS posts transaction for score.

Outcome: Millisecond JSON response with probability.