Travis Pettijohn - GitHub Portfolio
https://linkedin.com/in/pettijohn
REP JOT
- https://github.com/Pettibyte/RepJot
- Built a distraction-free, privacy-first weightlifting tracker that runs on constrained legacy browsers, including the Kindle Scribe experimental browser.
- Developed a Svelte/Vite single-page application compiled to ES2019, balancing modern UI development with legacy-device compatibility.
- Designed a zero-infrastructure architecture using Google OAuth, Google Drive app-data storage, and GitHub Pages hosting; workout data remains in each user’s Google account.
- Implemented three-way JSON merge and conflict handling to synchronize workout history across multiple devices.
- Built and evaluated an agent-assisted development workflow with Qwen-3.8 Flash Next, running locally on Strix Halo 128GB, using independent agents for implementation, code review, and targeted remediation of authentication and synchronization logic.
Rust Linux Kernel Driver for Corsair AI Workstation Performance Selector
- https://github.com/pettijohn/corsair-ai-workstation-performance-level-linux
- Developed a Rust Linux kernel module that integrates unsupported Corsair workstation firmware controls with Linux via ACPI/WMI and sysfs.
- Reverse-engineered WMI method/event payloads to decode real-time performance states: Quiet, Balanced, Max, and Super.
- Built an event-driven GNOME status indicator that blocks on kernel notifications instead of polling, using ~4.8 MB RSS and 0% idle CPU in validation.
- Added production-minded tooling: Secure Boot module signing, MOK enrollment flow, install/uninstall scripts, kernel Rust toolchain checks, and dev-containerized builds.
- Designed a small no_std-friendly Rust decode crate with testable firmware parsing logic separated from kernel integration.
Whisrs Voice Dictation
- https://github.com/y0sif/whisrs/pull/50
- Whisrs is a Linux-native voice dictation software in Rust with support for multiple speech recognition providers including cloud and local.
- Pull request added support for realtime dictation for OpenAI-compatible endpoints, namely Lemonade.
- Before this PR, the app supported realtime for OpenAI cloud dictation, or batch transcription (process the whole recording when recording stops) for OpenAI-compatible, or realtime with Whisrs-hosted whisper.cpp (CPU-only).
- Pull request refactored OpenAI realtime transcription to a shared component that OpenAI-native and OpenAI-compatible can use.
- Added over 100 unit tests.
- End result: realtime, fully-local, fully-private, GPU-accelerated voice transcription for Linux in Rust.
- Merged and released in v0.1.19.
TaskSchedulerEngine (2011 through 2026)
- https://github.com/pettijohn/TaskSchedulerEngine
- Built a lightweight C#/.NET scheduling engine with cron-like rules, second-level precision, timezone support, async callbacks, graceful shutdown, and retry behavior.
- Designed a fluent scheduling API plus cron parsing, giving consumers an ergonomic way to define recurring or one-shot tasks without heavyweight scheduler infrastructure.
- Implemented a high-performance rule evaluator that compiles schedules into bit masks for fast per-second matching with minimal allocations.
- Architected a thread-safe evaluation pump that dispatches matching tasks asynchronously, tracks running callbacks, isolates failures, and handles missed-second catch-up behavior.
- Modernized the project for .NET 8, 9, and 10 with cross-targeted automated tests covering parsing, concurrency, retries, shutdown, and scheduler edge cases.
Portfolio Drawdown Monte Carlo
- https://pettijohn.github.io/PortfolioDrawdownMonteCarlo/ and source code
- Algorithm computes probability that a given portfolio shape will last through retirement, when retiree is in drawdown phase and not earning income.
- Kept skills sharp, writing the algorithm in C#, Rust, Go, and TypeScript, plus a web basic UI. All but TypeScript used parallel programming on this CPU-bound algorithm.
- Wrote code by hand in 2022.
- In 2026, AI-assisted to run all in WASM in the browser, including parallel multi-threaded WebWorkers.