Skip to content
View MonsterAzi's full-sized avatar
  • Plano, TX

Organizations

@choccynix

Block or report MonsterAzi

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
MonsterAzi/README.md

Azazelle

Data Science student in Dallas (Plano, TX). I mostly spend my time breaking PyTorch optimizers, building frontends for local LLMs, and over-engineering solutions to petty arguments.

What I'm working on

The serious ML stuff:

  • C-Optim: Contributed to implementing cautious momentum optimizers. Figured out how to fix C-optim for Muon entirely by accident in my own sandbox.
  • mergekit: Contributed to the toolkit based on insights from reading lots of merging papers.
  • AlphaEvolveWriting: Worked on an evolutionary system that forces AI stories through ELO tournaments to improve them.

Local AI & Tools:

  • BunnyTavern: A fast, local-first web UI for AI roleplay built on Svelte 5 and Bun. Mostly built it because I want to experiment with it in the future.
  • SillyTavern-CharacterSimilarity: Uses KoboldCpp's embedding endpoints to run kNN outlier detection and semantic searches on characters. Honestly? My best project.

Cursed math & spite-driven development:

  • Gradient Descent Vis: A janky Matplotlib 3D widget I wrote specifically to watch Adam go the wrong direction in real-time.
  • Mario-Kart-Calculator: A linear weighted model built over a weekend purely to find what loadout is good. (It is way out of date now and is kind of oversimplified.)

The Stack

  • Daily: Python, PyTorch, Pandas, NumPy
  • Currently fighting with: Rust, Mojo, LaTeX

Azazelle's stats Top languages

Pinned Loading

  1. SillyTavern-CharacterSimilarity SillyTavern-CharacterSimilarity Public

    JavaScript

  2. BunnyTavern BunnyTavern Public

    A fast, local-first Svelte 5 web UI for AI character roleplay. Built on Bun and designed to plug directly into OpenAI-compatible backends (vLLM, Ollama) with native support for reasoning models.

    Svelte

  3. Mario-Kart-Calculator Mario-Kart-Calculator Public

    A vanilla JS weekend project to find the mathematically optimal Mario Kart World loadout using a linear weighted model.

    HTML

  4. kyleliang919/C-Optim kyleliang919/C-Optim Public

    [ICLR 2026] When it comes to optimizers, it's always better to be safe than sorry

    Python 415 14

  5. tobysimonds/AlphaEvolveWriting tobysimonds/AlphaEvolveWriting Public

    Python 120 11