I design and build interconnected software systems across hardware, operating systems, native interfaces, real-time audio and visuals, and data-processing infrastructure.
I work from the desired utility and real constraints: shaping the product, defining how its components interact, implementing a useful first version, and developing the architecture as requirements grow. That includes software-defined controls and visual feedback, compatibility layers, local-first applications, and scientific tools—not just connecting existing services.
My projects use Rust, Python, TypeScript, native platform frameworks, and C++ where an SDK or system boundary requires it. I investigate unfamiliar pieces, test the difficult interactions, and carry the work through to something another person can operate and maintain.
A native Linux VST3 proxy and supervised Windows plug-in host, built in Rust with C++ platform and SDK interfaces. The work covers cross-process audio and events, plug-in state and project recall, native editor lifecycles, and verified installation and recovery on Steam Deck with Bitwig.
Active development, with documented results for specific plug-ins and host configurations—not universal compatibility. The repository is publicly readable; its license remains proprietary.
A graphics companion for Ableton Push 3 in Controller Mode alongside DrivenByMoss and Bitwig. It combines hardware display control, a single display-writer owner, authenticated frame delivery, and explicit stale-frame and disconnect behavior. Native window-capture research connects the computer's visual output to the physical display while preserving ordinary musical controls.
Physical display and transport paths have been exercised. Selecting and qualifying the final capture implementation remains active work. This does not run Bitwig on the Push itself or modify its firmware.
Project and current boundaries
A Rust scientific data and execution platform, designed from the ground up to keep scientific definitions separate from execution machinery. Current work includes typed scientific composition, prepared project execution, bounded data movement, durable datasets with independent readback and reuse, resource accounting, and a native project Workbench.
The product goal extends from flexible research to preparing and executing useful applications, with deployment as a further target. GEP remains pre-production; general distributed execution and standalone deployment packaging are not delivered capabilities.
Selected source code, architecture documents, and central product-design dossiers are available for private technical walkthroughs with prospective clients and employers. Private application research is not part of the public portfolio.
A local-first Apple Calendar and Reminders board: signed macOS EventKit bridge, authenticated private delivery, SQLite mirror, and React/Fastify clients for iPad and Steam Deck. The read-only end-to-end path has been manually validated; unattended synchronization remains in development. The interface distinguishes current, stale, and unavailable data without writing back to Apple sources.
Two related projects address different parts of the device experience:
- SerialOSC for Steam Deck: a released rootless SteamOS installer around upstream serialosc, with user services, diagnostics, removal, and documented multi-device testing.
- PlugData Monome Devices: grid and arc discovery, device ownership, hot-plug recovery, and LED feedback for software-defined controls. Documented hardware and host tests distinguish passing paths from unresolved lifecycle cases.
Steam Deck installer and release · PlugData device interfaces
Sample and live-buffer playback work in PlugData: loop boundaries, direction changes, playback rate, switching, and mixer behavior. Development uses recorded audio comparisons and physical interaction, not just a passing build. Playback repairs are documented; recording and broader device integration remain separate work.
Application and development evidence
A live PDF-to-Markdown pipeline built around OpenDataLoader. It converts sample business PDFs and preserves useful document structure, including tables. Visitors can try the real conversion path rather than watch a simulated result.
Try the live application · View the source
I am available for native and cross-platform applications, hardware and protocol integration, real-time visual/audio tools, Rust data infrastructure, and turning technical prototypes into usable products. Python automation, API integration, and document processing remain part of that work.
Bring the intended result and the constraints. I can help define the interaction and architecture, deliver a useful first version, and develop it as the requirements become clearer—with tested boundaries and a practical handoff.