A small, honest collection of side projects and open-source contributions. Day-job work at Microsoft is in Experience.
Real-time charging-wattage display, in your macOS menu bar.
A lightweight macOS Menu Bar app that shows the live wattage your laptop is pulling from its charger. Built with Swift and AppKit, polls `ioreg` with configurable cadence (1 / 5 / 10 / 30 s), reacts instantly to plug/unplug events, and can hide itself when running on battery.
Natural language → terminal commands. Just ask.
A Rust CLI that turns plain English into a shell command, shows it to you, and asks before running. Tested with Qwen 3.5-9B and GPT-5.2 on macOS, with CI-validated Linux and Windows builds and prebuilt binaries on Releases.
Fast, single-binary CLI for generating images from text prompts.
A Rust CLI that turns text prompts into images, speaking multiple providers natively — switch between OpenAI, Azure OpenAI, Azure MAI, Google Gemini, and LocalAI by changing one config value. Built on a synchronous HTTP stack (`ureq` + `rustls`) for fast cold starts and a small binary, with an interactive setup wizard, shell completions, and prebuilt binaries on Homebrew, crates.io, and Releases.
Local camera feeds for development — MJPEG, RTSP, optional YOLO overlay.
A small Python tool (on PyPI) that exposes your webcam as an HTTP MJPEG or RTSP stream from a single command. Optional `[vision]` extra adds a built-in YOLOv8 human-detection overlay — handy when you need a controlled camera source for testing CV pipelines.
A Foundation Model for Multimodal AI Agents (CVPR 2025) — added the FastAPI inference server.
Contributed a FastAPI-based REST inference server to the official Microsoft Magma repository (CVPR 2025), enabling the multimodal model to be served via HTTP endpoints for vision + language requests. The contribution included the server implementation, deployment options, dependency-install fixes, and README updates.