PhD student in EECS. AI and quantum annealing for power systems and power electronics.
I've been working with AI since 2020 and never got around to making a GitHub for myself, so here I am. Below are some tools for Claude Code and a few fun apps. I built all of them around one habit I've kept since the first LLMs: verify what a chat model would otherwise just assert.
|
research-hound A five-stage research workflow for Claude Code, with 21 scripts that check citations and claims instead of trusting the model. |
skill-vision Validates Claude Agent Skills for spec conformance, quality and security, and counts the context tokens each one costs. |
agent-oracle Decides how an agent system should be built before it runs, from the number of agents to what gets locked down. |
|||
|
mr-usage A Rust and Tauri desktop app that shows what is left on nine paid APIs, with keys kept in the OS keystore. |
llm-dogfight Two small local language models command rival saucer fleets in your terminal while a genetic algorithm evolves their tactics. |
OpenCarMind A car logger with an on-chip neural network for the ESP32-S3 and the early-access ESP-Mosaico (ESP32-S31), plus a desktop app for Windows, macOS and Linux. |
NVIDIA SkillSpector scans the three skills on every push, and each ships with a .skillspector-baseline.yaml that names every suppressed finding and why.
One-time prices in US dollars, before sales tax. All four are on sale through Gumroad, with a 14-day refund on every pack.
| GlassVGL.comp Liquid glass for the LVGL app you already have |
$3.99 | |
| GlassVGL Icons A frosted-glass icon theme for Linux desktops, in light and dark |
$8.99 | |
| GlassVGL Edge GUI The LVGL glass theme, royalty-free in your own products |
$12.99 | |
| GlassVGL Desktop Frosted glass for your whole GNOME desktop, light and dark. Linux only. |
$19.99 |
Free samples for every toolchain, seven app screens and a 25-icon sample of GlassVGL Icons are in the public GlassVGL repository.
Four of my own, still in active development.
| LightEDA A Rust and Tauri desktop IDE, now in alpha, that builds and flashes ESP32 and Arduino firmware and runs the open-source FPGA flow, with an on-device assistant. The Pro waitlist is open. |
|
| Murasaki Invite only. A purple-team platform for authorized security research, with AI-driven reverse engineering and ICS threat modeling. |
|
| OpenCarMind A car logger with an on-chip neural network for the ESP32-S3 and the early-access ESP-Mosaico (ESP32-S31), plus a desktop app for Windows, macOS and Linux. |
|
| TigerOCR Hosted GPU OCR that turns PDFs and images into clean Markdown, as a remote MCP server and REST API. |
I'm on Claude's Ultra plan, with advanced agent access in Claude Duet. Duet runs as the orchestrator and sends work out to many Claude agents in parallel across my projects.
- π Each agent logs its progress in detail. What reaches me is results, blockers and questions, never status chatter.
- π No research claim gets trusted until hall-hunter has been over it.
- π Every change goes in as a pull request and gets reviewed before it merges.
- π§ CI stays lean: quick Linux checks on pull requests, full builds only for releases.
I also have Gemini and Perplexity subscriptions. Both are good, but I open them far less, because I think Claude is the best.
Outside the tooling, my research is AI and quantum annealing, applied to power systems and power electronics.
My Claude Code sessions run in the cloud with a short rulebook that Claude reads before anything else.
- π Anything that touches the live web goes through agent-reach: SerpApi answers which pages rank, Firecrawl reads the page, and its developer, research and government indexes handle code, papers and law. The built-in search stays off unless I ask for it, because I want results I can audit.
- βοΈ Every draft that leaves a session passes through humanizer, and anything that looks too confident gets hall-hunter, which has to name the file, line and contradicting fact for every finding or report it as unverified.
- β Skills get validated by skill-vision before they ship. Agent designs start with agent-oracle's four questions. Papers go through research-hound, which checks that each DOI resolves and that its metadata matches the citation.
- βοΈ The quantum-computing skill covers Qiskit, PennyLane, QuTiP and D-Wave Ocean, which is where the power systems work and the Claude work meet.
Everything on GitHub stays free. If it saved you time, these are the ways to give some back, as listed on buildforpeople.dev/support.
- π GitHub Sponsors: opening soon. Until then, following me on GitHub or starring a repository helps.
- ποΈ Buy a pack: sales pay for the free versions. See Products above.
- β Free: star a repository, share a project with someone who would use it, or open an issue when something breaks. Bug reports with steps to reproduce are worth a lot.
- βοΈ Something else in mind? Questions, or want to sponsor a feature for your lab or company? Email support@buildforpeople.dev.








