reg-change-engine

Problem

Regulatory teams need to stay on top of new government rules every day. Reading the full feed isn't realistic.

What I built

A model trained on 8K labeled rule excerpts, wrapped in a web dashboard with scheduled updates.

Outcome

89% accuracy and fast enough to use every day, turning a long legal feed into a short list of what to read first.

What I learned

• Built the model and the tool around it, not just the model on its own.
• Tuned for speed so it can actually be used in a daily workflow.