Nate Williams / Open Source Contributor
OpenCode

48x Performance Fix

OpenCode's diff view was spawning 2n git processes for n files, each with 10-20ms of overhead. For 1,000 files, that meant a 14-second freeze every time.

I batched all file reads into a single git cat-file --batch command, cutting 2,000 process spawns down to 1.

14s 0.29s diff calculation for 1,000 files
View Pull Request ↗

"I appreciate the hard work @natewill"

Kit Langton, Engineer at OpenCode

Workspace Creation: 5x Speedup

14s ~3s workspace creation

Creating a new workspace in large codebases froze the app for 12-15+ seconds. A race condition between worktree creation and client bootstrap caused thousands of file watcher events to fire while the worktree was still being populated. I added a pending state that delays bootstrap until creation is actually complete.

Go Performance Optimization

2am Bug Fix

OpenCode's desktop app went down late at night and no one on the team was awake. I dug into the issue, found a DevTools workaround to unblock users, and posted it to the Discord and GitHub.

GitHub comment with workaround, 10 reactions
Sergio: Thank you so much for this. I came here to ask advice for this issue too.
Anguzo: @natewill Thanks! This temp fix works
O_meta: 谢谢!
ДАРОВА: its working

OpenCode

Open Source Contributor
Present
  • Contributing to an open-source startup product serving 6.5M active users, working directly in a real production codebase.
  • Implemented a performance optimization in the Tauri desktop app (Next.js, React, TypeScript) that made diff calculation 48x faster and worktree creation ~5x faster.
  • Built BetterPRs, a full-stack pull-request filter tool (Next.js, PostgreSQL, Gemini API) to help maintainers manage incoming PRs more efficiently.
  • Investigated live user-reported bugs from Discord and GitHub, reproduced issues from incomplete information, and delivered fixes or workarounds that unblocked users in real time.
Go TypeScript React Next.js Tauri

Department of Defense x Virginia Tech

Undergraduate AI Researcher
Aug 2024 – June 2025
  • Supported a DoD research effort focused on AI-driven automation for converting an outdated systems-engineering modeling language to the updated SysML standard.
  • Explored LLM fine-tuning, open-source models, and benchmarking workflows using commercial and open-source model families under strict data-security constraints.
Python PyTorch LLM Fine-Tuning SysML

InteriorAI

Founder / Builder
Summer 2025
  • Founded and built an AI-powered interior design application exploring image generation and furniture-visualization workflows.
  • Built a full-stack application using Next.js, Vercel, Gemini API, and segmentation models to prototype room visualization and furniture-placement workflows.
  • Built and iterated on several MVPs, using direct user feedback to test demand and refine the product concept.
Next.js TypeScript Gemini API Vercel

InteriorAI

AI-powered interior design tool with three modes: add furniture to a space, transform full room styles, and identify furniture from generated images. Uses SAM3 computer vision and fal.ai for image generation with Three.js visualization.

Next.js TypeScript Three.js fal.ai SAM3

BetterPRs

PR filter for open source maintainers. Uses LLMs to score pull requests by review-worthiness based on size, community engagement, and review difficulty. Configurable scoring weights and filters by date, scope, and type.

Next.js TypeScript Drizzle ORM Trigger.dev Sentry

Mussel Counter

VT capstone project for the Department of Fish and Wildlife Conservation. Automates counting and classification of juvenile freshwater mussels in microscope images using Fast R-CNN, replacing tedious manual inspection with a human-in-the-loop desktop application.

React Electron Python FastAPI Fast R-CNN SQLite
Nate Williams

Senior at Virginia Tech studying Computer Science, graduating Spring 2026.

Education B.S. Computational Modeling & Data Analytics
Minors Computer Science, Mathematics, Statistics
Expected May 2026