Python/C++/PyTorch/OpenCV/ONNX - Represented the company in the CVPR 2023 Face Anti-Spoofing Challenge, achieving 9th place out of ~50 companies. - Co-developed a 1:1 Face Verification API for the NIST FRTE 1:1 Verification, achieving 78th place out of 577 submissions in non-standardized facial image recognition, significantly improving the company's ranking. - Developed a cross-platform Face Recognition SDK in C++, supporting Windows, Linux, and Android with GPU and SoC hardware acceleration. Led the design and optimization of the SDK’s architecture, focusing on face detection and anti-spoofing model development and deployment. The SDK is used in the company’s Android-based access control devices and server-side recognition systems. - Deployed face-related AI models on CPU/GPU/SoC using frameworks such as ONNXRuntime, OpenCV DNN, and TensorFlow Lite, across Linux and Android platforms. - Established data annotation workflows for face and facial landmark detection. - Developed lightweight RGB or RGB-D models for multi-attribute facial prediction. - Co-organized a self-supervised learning study group, presenting research on Masked Image Modeling (MAE and ConvNeXt V2) and self-supervised learning in face analysis.

Python(PyTorch)/R - Developed an advanced image feature extraction method using PyTorch, capturing 14x more features compared to traditional human-defined datasets, tailored to biological research needs. - Designed and implemented custom neural networks for supervised and unsupervised tasks, building training pipelines and optimizing models. - Applied Explainable AI (XAI) techniques (e.g., Grad-CAM) to enhance model interpretability and generate visual insights. - Led an internal AI study group, sharing knowledge on self-supervised representation learning and CycleGAN with a team of 7 participants, facilitating knowledge transfer and fostering team collaboration on advanced machine learning topics. - Conducted exploratory data analysis using Matplotlib, NumPy, and Pandas; performed statistical analysis of visual features with regression models and diversity metrics in R. - Co-authored a scientific manuscript detailing the proposed biological image analysis approach in collaboration with the Principal Investigator and research team.

C++/R/Python - Developed a custom social evolution model in C++ to simulate eco-evolutionary dynamics, exploring complex interactions between sociality, environmental quality, and population dominance. - First author on a paper published in Frontiers in Psychology (2022) and co-first author on a paper in Ecology Letters (2023), collaborating closely with the Principal Investigator and research team.