AI And .NET: Building Intelligent Apps with .NET 9

The combination of artificial intelligence (AI) and modern software development is changing the way we build applications. With the release of .NET 9, Microsoft keeps giving developers tools and frameworks that make it easy to add AI capabilities to strong, scalable applications.

In this article, we’ll look at how .NET 9 helps with AI-driven development, focusing on how to use machine learning models and walk through a sample project—an intelligent sentiment analysis API—complete with example code.

Why .NET 9 for AI Development?

.NET 9 has some great new features, like better performance, better support for cloud-native apps, and a refined ecosystem that works really well with AI workloads. It’s got APIs that are super simple, better AOT (Ahead-of-Time) compilation, and better integration with containers, so it’s a great choice for building intelligent, high-performance apps. On top of that, libraries like ML.NET and compatibility with popular AI frameworks like TensorFlow and ONNX let developers embed AI right into their .NET solutions.

Setting the Stage: Tools and Prerequisites

To follow along, you’ll need,

• Visual Studio or VS Code with .NET 9 SDK: Install the latest version from the official Microsoft site, for coding and debugging. I will be using VS Code for the demo project.
• ML.NET: A machine learning framework for .NET (install via NuGet: Microsoft.ML).
• A Basic Understanding of ASP.NET Core: For building the API.

Project Overview: Sentiment Analysis API

Let’s build a simple ASP.NET Core Web API that uses ML.NET to do sentiment analysis on text submitted by users. The API will predict if a text expresses positive or negative sentiment, which is a great starting point for AI-driven apps.

Complete Source code: Click here

Step 1. Setting Up the Project.

Create a new ASP.NET Core Web API project.

dotnet new webapi -n dotnet9SentimentApi -f net9.0
cd dotnet9SentimentApi

Open the folder in VS Code. The project should be as shown below.

Let’s add the ML.NET NuGet package.

You can either run the command or do it from VS code UI as illustrated below.

dotnet add package Microsoft.ML

Step 2. Define the Data Models.

Create a folder named Models and add two classes: SentimentData and SentimentPrediction.

SentimentData.cs: Represents the input data for the model.

namespace dotnet9SentimentApi.Models;

public record SentimentData
{
    public string Text { get; set; } = string.Empty;
}

SentimentPrediction.cs: Represents the model’s output.

namespace dotnet9SentimentApi.Models;

public record SentimentPrediction
{
    public bool Prediction { get; set; } // True = Positive, False = Negative
    public float Probability { get; set; }
}

Step 3. Training a Simple Sentiment Model with ML.NET.

For simplicity, we’ll use a pre-trained model approach here, but ML.NET allows you to train your own model with a dataset. Create a SentimentModelTrainer.cs class in a Services folder to simulate model setup:

To train the model with your own data.

private void TrainModel()
    {
        // Simulated training data (in practice, load from a file or database)
        var data = BuildTrainingData();

        var dataView = _mlContext.Data.LoadFromEnumerable(data);

        // Enhanced pipeline with text preprocessing
        var pipeline = _mlContext.Transforms.Text.FeaturizeText("Features", new Microsoft.ML.Transforms.Text.TextFeaturizingEstimator.Options
        {
            WordFeatureExtractor = new Microsoft.ML.Transforms.Text.WordBagEstimator.Options(),
            StopWordsRemoverOptions = new Microsoft.ML.Transforms.Text.StopWordsRemovingEstimator.Options()
        }, nameof(SentimentTrainingData.Text))
            .Append(_mlContext.BinaryClassification.Trainers.LbfgsLogisticRegression(labelColumnName: nameof(SentimentTrainingData.Sentiment)));

        // Train the model
        _model = pipeline.Fit(dataView);

        // Optional: Save the model for reuse
        _mlContext.Model.Save(_model, dataView.Schema, "sentiment_model.zip");
    }

Note. In a real-world scenario, you’d train with a larger dataset (e.g., from a CSV file) and save the model using _mlContext.Model.Save().

Complete SentimentModelTrainer.cs class

using System;
using dotnet9SentimentApi.Models;
using Microsoft.ML;

namespace dotnet9SentimentApi.Services;

public class SentimentModelTrainer
{
    private readonly MLContext _mlContext;
    private ITransformer _model;

    public SentimentModelTrainer()
    {
        _mlContext = new MLContext();
        TrainModel();

    }

    private void TrainModel()
    {
        // Simulated training data (in practice, load from a file or database)
        var data = BuildTrainingData();

        var dataView = _mlContext.Data.LoadFromEnumerable(data);
        // Enhanced pipeline with text preprocessing
        var pipeline = _mlContext.Transforms.Text.FeaturizeText("Features", new Microsoft.ML.Transforms.Text.TextFeaturizingEstimator.Options
        {
            WordFeatureExtractor = new Microsoft.ML.Transforms.Text.WordBagEstimator.Options(),
            StopWordsRemoverOptions = new Microsoft.ML.Transforms.Text.StopWordsRemovingEstimator.Options()
        }, nameof(SentimentTrainingData.Text))
            .Append(_mlContext.BinaryClassification.Trainers.LbfgsLogisticRegression(labelColumnName: nameof(SentimentTrainingData.Sentiment)));

        // Train the model
        _model = pipeline.Fit(dataView);

        // Optional: Save the model for reuse
        _mlContext.Model.Save(_model, dataView.Schema, "sentiment_model.zip");
    }

    public SentimentPrediction Predict(string text)
    {
        var predictionEngine = _mlContext.Model.CreatePredictionEngine<SentimentData, SentimentPrediction>(_model);
        return predictionEngine.Predict(new SentimentData { Text = text });
    }

    //build training data
    private List<SentimentTrainingData> BuildTrainingData()
    {
        return new List<SentimentTrainingData>
        {
            new() { Text = "I love this product!", Sentiment = true },
            new() { Text = "This is terrible.", Sentiment = false },
            new() { Text = "The weather is nice!", Sentiment = true },
            new() { Text = "Horrible service, never again.", Sentiment = false },
            new() { Text = "Absolutely fantastic experience!", Sentiment = true },
            new() { Text = "It’s a complete disaster.", Sentiment = false },
            new() { Text = "I’m so happy with this!", Sentiment = true },
            new() { Text = "Disappointing and awful.", Sentiment = false },
            new() { Text = "I’m so impressed!", Sentiment = true },
            new() { Text = "I’m never coming back.", Sentiment = false },
            new() { Text = "I’m so excited!", Sentiment = true },
            new() { Text = "I’m so disappointed.", Sentiment = false },
            new() { Text = "I’m so pleased with this!", Sentiment = true },
            new() { Text = "I’m so upset.", Sentiment = false },
            new() { Text = "I’m so satisfied with this!", Sentiment = true },
            new() { Text = "I’m so angry.", Sentiment = false },
            new() { Text = "I’m so grateful for this!", Sentiment = true },
            new() { Text = "I’m so annoyed.", Sentiment = false },
            new() { Text = "I’m so thankful for this!", Sentiment = true }
        };
    }

}
public record SentimentTrainingData
{
    public string Text { get; set; } = string.Empty;
    public bool Sentiment { get; set; }
}

Step 4. Build the API Controller.

Create a SentimentController in the Controllers folder.

using dotnet9SentimentApi.Models;
using dotnet9SentimentApi.Services;
using Microsoft.AspNetCore.Http;
using Microsoft.AspNetCore.Mvc;

namespace dotnet9SentimentApi.Controllers
{
    [Route("api/[controller]")]
    [ApiController]
    public class SentimentController : ControllerBase
    {
        private readonly SentimentModelTrainer _modelTrainer;

        public SentimentController(SentimentModelTrainer modelTrainer)
        {
            _modelTrainer = modelTrainer;
        }

        [HttpPost("analyze")]
        public ActionResult<SentimentPrediction> AnalyzeSentiment([FromBody] SentimentData input)
        {
            if (string.IsNullOrEmpty(input.Text))
                return BadRequest("Text cannot be empty.");

            var prediction = _modelTrainer.Predict(input.Text);
            return Ok(new { Sentiment = prediction.Prediction ? "Positive" : "Negative", Confidence = prediction.Probability });
        }
    }
}

Step 5. Register Services in the Program.cs.

Update Program.cs to use minimal APIs and register the service.

// Add services to the container
builder.Services.AddControllers();
builder.Services.AddSingleton<SentimentModelTrainer>();

// Configure the HTTP request pipeline
app.UseHttpsRedirection();
app.MapControllers();

Complete the code for the Program.cs

using dotnet9SentimentApi.Services;

var builder = WebApplication.CreateBuilder(args);

// Add services to the container.
// Learn more about configuring OpenAPI at https://aka.ms/aspnet/openapi
builder.Services.AddOpenApi();

// Add services to the container
builder.Services.AddControllers();
builder.Services.AddSingleton<SentimentModelTrainer>();

var app = builder.Build();

// Configure the HTTP request pipeline.
if (app.Environment.IsDevelopment())
{
    app.MapOpenApi();
}

// Configure the HTTP request pipeline
app.UseHttpsRedirection();
app.MapControllers();

app.Run();

Our solution is ready to test.

The folder structure looks as shown below.

Test the API

dotnet run

You can use Postman to test the API or HTTP file in VS Code. I will test using VS Code .http file.

@dotnet9SentimentApi_HostAddress = http://localhost:5288

# post analyze sentiment
POST {{dotnet9SentimentApi_HostAddress}}/api/sentiment/analyze
Content-Type: application/json

{
  "text": "I am not happy with the result!"
}

###

The result is shown below.

The result is not as expected – indicates an issue with the sentiment analysis model’s training or configuration. Given the tiny dataset and simplistic setup in the original sample, the model isn’t learning meaningful patterns.

However, we can improve the result with more training data and testing.

Please find the complete source code: click here

Enhancing the Project

1. Scalability: Deploy to Azure with container support for cloud-native scaling.
2. Model Accuracy: Use a real dataset (e.g., IMDB reviews) and train with more sophisticated algorithms like FastTree.
3. Performance: Cache predictions for frequently analyzed text using MemoryCache.
4. Integration: Extend with Azure Cognitive Services or TensorFlow.NET for advanced AI capabilities.

Conclusion

.NET 9, paired with ML.NET, offers a powerful platform for building intelligent applications. This sentiment analysis API demonstrates how easily you can integrate AI into your .NET projects. As AI continues to evolve, .NET’s robust ecosystem ensures developers can stay at the forefront of innovation. Experiment with this project, explore .NET 9’s new features and unlock the potential of AI in your next application.