The work

Things I have built.

Six of them are written up in full: what the problem actually was, what I chose and why, and the number it moved. The rest of the list is here too, with links to the public record.

  • 40+hours a week returned by one agent
  • 85%fewer errors on the quotes it reviews
  • 75 to 95%customer satisfaction after AI coaching
  • 0tokens per recall, memory built outside the model

Open line

Building something like this? I like these conversations.