Skip to content
View soyuz43's full-sized avatar

Block or report soyuz43

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 supported. This note will be visible to only you.
Report abuse

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

Report abuse
soyuz43/README.md

William Stetar

Building tools and frameworks to understand how language models fail—and why it matters for alignment.

AI Safety Researcher | Computational Linguist | Systems Developer

I analyze language model behavior through computational linguistics, focusing on how discourse patterns shape alignment failures. I also build tools for LLM integration, git automation, and symbolic reasoning systems. Current Research: Investigating epistemic failure modes in LLMs using Systemic Functional Linguistics—specifically how models prioritize interpersonal coherence over factual accuracy under pressure. Technical Skills: Go, Python, Bash, C#, SQL | LLM integration | CLI tooling | Git automation

Projects
  • PRbuddy - Auto-generates pull request drafts using Git hooks and LLM infrastructure

  • Symbolic Grammar Interpreter - Recursive system for analyzing contradiction and structural drift in text

  • Holoplan CLI - Transforms user stories into Draw.io wireframes via multi-agent LLM reasoning

Writing & Research

📝 Hashnode Blog - Essays on AI safety, linguistic analysis, and ethical implications of LLMs

Notable pieces:

  • "The Nuremberg Defense of AI" - On accountability in ML development

  • "Epistemic Autoimmunity" - Framework for analyzing alignment failures through discourse patterns

Currently
  • Conducting comparative analysis of Western vs. Eastern LLM failure modes
  • Developing empirical framework for measuring epistemic autoimmunity in model outputs
  • Open to collaboration and full-time research positions in AI safety
Contact

📧 kebekad673@proton.me 🔗 LinkedIn

Pinned Loading

  1. PRbuddy PRbuddy Public

    Auto-generate pull request drafts and explain commits using Git hooks + LLMs

    Go

  2. ftype-audit-safe ftype-audit-safe Public

    Safe PowerShell tool for auditing and repairing Windows file type associations with backup, dry-run, and signature support.

    PowerShell

  3. holoplan-cli holoplan-cli Public

    A CLI tool that transforms user stories into Draw.io wireframes via multi-agent LLM reasoning. Chunks narratives into views, builds layouts, validates geometry, and outputs clean XML. Tech stack: G…

    Go

  4. DreamTracker DreamTracker Public

    DreamTracker is a full-stack web application for recording, managing, and reflecting on dreams. With features like tagging, favoriting, keyword filtering, and client-side caching via React Query, i…

    C#