Why? I teach IT professionals who live in PowerShell. Python ML examples don't resonate. Building from scratch teaches fundamentals better than using black-box libraries.

What's included:

- Neural networks: Backpropagation, gradient descent, multiple activation functions - Q-learning agents: Experience replay, epsilon-greedy exploration, reward shaping - Environment simulation: Market competition, resource optimization - Real-time visualization dashboards

Technical details:

- Pure PowerShell (classes, no C# interop) - ~3000 lines of code - Compatible with PowerShell 5.1+ (Windows/Linux/Mac) - Available on PowerShell Gallery

Example use case: Q-learning agent learns to build ASCII castles through trial-and-error. Agent receives aesthetic rewards and discovers patterns without hardcoded rules.

Practical applications: - IT automation that adapts to patterns - Predictive maintenance from log analysis - Resource scheduling optimization - Anomaly detection in enterprise systems

I know this isn't production ML (Python/PyTorch are better for that), but it's valuable for: 1. Teaching ML concepts to non-Python developers 2. Understanding algorithms by implementing them 3. PowerShell-native automation that learns

GitHub: https://github.com/JupyterPS/VBAF PowerShell Gallery: Install-Module VBAF

Happy to answer questions about the implementation or design decisions.