ML Systems / Computer Engineering; Embedded Systems / Hardware–Software Integration; GPU Computing / Vulkan / PyTorch Systems. Supporting work spans Computer Vision, Full-stack, and AI Agents.
- MIRA: 90.6% and 90.58% mAP50 on the 415-image validation split; these are validation results, not real-world sorting accuracy. The independent 1,375-image test split was not evaluated.
- FluidicStudio: core software workflows implemented; pump-driver detection is experimental.
- ESP32-S3 Alarm Clock: basic serial testing is documented; full-board bring-up is not.
- Poorup: authoritative Node.js/Socket.IO multiplayer server with reconnect recovery, automated tests and CI.
TorchVK is early research; it does not yet provide a complete PyTorch backend.


