Challenge
The original Python desktop application for contactless microwave radar vital-sign monitoring struggled with UI frame drops, GIL contention, and packaging fragility when processing real-time IQ radar streams alongside CNN inference.
VitalMonitor RS: Contactless Microwave Radar Vital Signs
Rewriting a legacy Python medical radar workstation into a hardware-accelerated Rust, GPUI, and ONNX real-time telemetry system.
- Rust
- GPUI
- ONNX Runtime
- Digital Signal Processing
- Embedded & Medical IoT
Confidential
Approach
We executed a complete Rust rewrite organized as a multi-crate workspace combining bandpass DSP filtering, FFT frequency analysis, envelope detection, and dynamically loaded ONNX Runtime CNN models for heart-rate (40-120 BPM), breathing-rate (10-30 BPM), and patient-state classification, rendered at 60fps in Zed's GPUI framework.
Outcome
Replaced a fragile Python runtime with a deterministic native binary capable of real-time multi-algorithm fusion and synthetic IQ sensor simulation without hardware attached.
Results
- Complete Rust + GPUI rewrite of the legacy Python radar monitoring suite
- Multi-algorithm fusion combining FFT, envelope detection, and ONNX CNN inference
- Built-in synthetic IQ radar waveform generator for hardware-free validation