Tom Compagno
  • Special Series
  • Blog
  • About
Get started

Tagged Articles

Training

3 posts

Poisoned Wells – Training Data Flaws

A look at the root cause of bias: how incomplete or historical datasets (like 50 years of male-dominated resumes) "teach" the AI to be discriminatory.

Learn more

The Guardrails – RLHF (Human Feedback)

The final "polish" where humans grade AI answers to make them safer, more accurate, and less "robotic."

Learn more

The Training Phase – Pre-training & Fine-Tuning

How an LLM goes from reading the entire internet to being a helpful assistant that follows instructions.

Learn more
  • Home
  • Special Series
  • Blog
  • About
  • Moira's Lexicon
Tom Compagno

Tom built this site in Webflow, then migrated to Astro

  • Privacy
  • Contact
  • LinkedIn
All rights reserved © 2026
Made partially by AI