SlimeVR is an open source organization for full-body motion capture in VR. It uses IMUs and a forward kinematics model to allow social VR users to have legs in games like VRChat. I am a core contributor in the org, and the primary author of several components of the ecosystem.

The four main things I work on under this umbrella, are located in the SlimeVR-Rust monorepo or in SolarXR.

I develop bare metal embedded rust on the ESP32C3 and NRF52840 microcontrollers, which takes IMU sensor data, does sensor fusion, and sends that data over WiFi or BLE. I use async/await and embassy for concurrency without the need for an RTOS¹.

This firmware has been really fun to work on and is what I spend most of my time on right now.

If that sounds exciting to you, open a PR or support my work so I can buy more hardware :)

I also am the primary author of an OpenVR overlay for SlimeVR, which renders a simple 3D view in VR of the pose of the user. This has been critical in debugging SlimeVR, and allows power users to tweak their settings in real time.

I also created ovr_overlay, which allows rust code to access the C++ based OpenVR API. Prior to this work, only fairly out of date and unmaintained bindings to the C API existed. I used autocxx to easily write FFI for the C++ code.

I created an experimental skeletal model that can solve the user's pose. The goal is to be more general purpose than the current (trigger warning) Java based model that the official server uses, with several new features like arbitrary positional constraints, adjusting mounting calibration without a full reset, and supporting languages other than Java, via FFI.

Unfortuantely, a lot of the skeletal model is still feature incomplete. If you are interested in helping me build this, reach out to me on the SlimeVR discord.

I maintain and co-authored the SolarXR protocol, which is a Websocket and Flatbuffer based networking protocol that allows applications (like the GUI and overlay) to work with the SlimeVR server.

After we adopted the protocol, it allowed the team to break the gui code into its own app and adopt technologies like tauri, typescript, and react, with clearer separation of concerns, less code, and prettier UI/UX.