You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Scenario Unlocked
unlocks solutions where human judgement and decision-making is involved. Sufficiently important decisions require deeper analysis. If a quantitative approach is available, it is usually the one preferred as it will offer the best combination of an approach and outcome.
Baseball AI Workbench
is a web application that showcases performing quantitative decision analysis (decision thresholding, what-if analysis, AI Agents with probability & confidence interval analysis) using in-memory Machine Learning models with historical baseball data.
The application has the following features:
Historical position player (batters) up to the end of the 2024 season
Three different decision analysis mechanisms to perform what-if analysis
Agentic AI integrations with Agents performing research & quantitative analysis
A simple "expert" rules engine to predict baseball hall of fame induction, contrasted with a Machine Intelligence solution
Single and multiple machine learning models working together to predict baseball hall of fame ballot and induction probabilities
Machine Learning models are surfaced via ML.NET in-memory for rapid inference (predictions)
Surfaced via the Aspire.NET integration with a Blazor application framework using SignalR to deliver the predictions from the server to the web client at scale
Self-contained application with Docker, allowing you to run locally
Architecture - Cloud Deployment Diagram:
Project Structure (Verified):
Visual Studio 2022, .NET 9, Server-Side Blazor, ML.NET v4.02, Semantic Kernel, Azure AI Foundy, Azure OpenAI Azure SignalR (optional for massively scaling message communication for Azure deployments)
