I'm a mathematician and software engineer with 15 years of industry experience, doing independent research on LLM alignment and interpretability. I have a minor in philosophy.
I created Heretic, by far the most widely used software for removing censorship/safety alignment from language models. More than half of all abliterated models ever published on Hugging Face were made using Heretic, and Heretic models have millions of downloads combined. Many of the top-ranked models on the UGI Leaderboard were created with Heretic.
I also designed the DRY and XTC token sampling algorithms, which are now implemented in most LLM inference engines, and are widely used for creative writing and roleplay applications.