Why should you care?

AIPredict ops needs live fraud catch rate and forecast error, not only raw logs.

See it live — copy this example

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

var metrics = JsonSerializer.Deserialize<ModelMetrics>(File.ReadAllText("models/fraud-v12.metrics.json"));
var dashboard = new {
    metrics.AreaUnderRocCurve,
    metrics.F1Score,
    DailyScores = db.Predictions.Where(p => p.Day == DateOnly.FromDateTime(DateTime.UtcNow))
        .GroupBy(p => p.Bucket).Select(g => new { g.Key, Count = g.Count() })
};
return Results.Ok(dashboard);

What happened?

• Store offline metrics JSON with each zip; query prediction tables for online aggregates feeding BI.
• Follow the steps below — typing the code yourself is the fastest way to learn.

Practice next

1. Save metrics.json on train.
2. Aggregate predictions by day.
3. Expose read API for BI.
4. Add challenger comparison chart.
5. Alert when catch rate drops.

Remember

Offline metrics file. Online aggregates. BI-friendly API.

AIPredict ops dashboard

Ops sees fraud catch rate vs last week.

Outcome: Incident opened before customers notice.