brand image

Ali

  • Publications
  • About
  • Blog
  • Recent
  • A recipe for measuring bias in healthcare AI
  • Getting the hang of instruction tuning

built with Hugo and poison
© Muhammad Ali 2025 . All rights reserved.

A recipe for measuring bias in healthcare AI

November 15, 2025
  • language models
  • evals
  • synthetic data
A step-by-step recipe for building an LLM judge to identify bias and equity-related harms in healthcare AI applications.

Getting the hang of instruction tuning

November 25, 2024
  • language models
  • tutorial
A hands-on programming tutorial of instruction tuning: I take a base Gemma 2B model and fine-tune it on the Alpaca dataset on a small GPU; this enables the model to follow user instructions.