Worked on the openvinotoolkit/openvino.genai repository to optimize image encoding within the VisionEncoderQwen2VL model. Refactored the preprocessing pipeline by integrating resizing, normalization, and reshaping operations directly into the OpenVINO model graph, enabling hardware acceleration and reducing unnecessary data movement. This approach improved throughput and lowered latency for image encoding, enhancing end-to-end inference performance for downstream computer vision workflows. The work demonstrated strong proficiency in C++ development, computer vision, and model optimization, with a focus on leveraging OpenVINO’s capabilities for efficient model-graph integration. No major bugs were addressed during this period, with efforts concentrated on feature development.