Akeso Eyecare
Android Staff Engieer
Ported gaze estimation deep learning models from Python to production C++ pipelines on Windows and Android, enabling real-time inference on edge devices.
Achieved 4x performance... more improvement on deep learning models through quantization, operator fusion, and ONNX Runtime optimization — without sacrificing accuracy.
Designed a robust on-device calibration pipeline with custom algorithms, improving per-user gaze accuracy and cutting setup time.
Built an on-device fine-tuning system in C++ using ONNX Runtime, enabling per-user model personalization without server dependency.
Shipped cross-platform native SDKs (Windows C++, Android NDK) consumed by downstream product teams.
Achieved 4x performance... more improvement on deep learning models through quantization, operator fusion, and ONNX Runtime optimization — without sacrificing accuracy.
Designed a robust on-device calibration pipeline with custom algorithms, improving per-user gaze accuracy and cutting setup time.
Built an on-device fine-tuning system in C++ using ONNX Runtime, enabling per-user model personalization without server dependency.
Shipped cross-platform native SDKs (Windows C++, Android NDK) consumed by downstream product teams.
