Experience

GitHub activity

Edinburgh University Formula Student

Developing autonomous racing vision pipelines for cone detection, fusing LiDAR and camera data for FSUK competitions.
Building a ZED stereo-camera inference pipeline, running detection models through ONNX Runtime and TensorRT for real-time performance.
Created and labelled a custom cone dataset in CVAT, establishing a reusable annotation pipeline for the perception team.
Fine-tuned and benchmarked YOLOv11 and RF-DETR across detection and segmentation metrics.

Southeast University — Summer AI Program

Proposed SparrowNet, an ensemble network combining lightweight ResNet variants with 1×1 convolutions.
Benchmarked it against ResNet18 and GoogLeNet V2 on an open Egyptian-hieroglyph dataset, analysing accuracy and efficiency.
Wrote up the methodology and results as a formal research paper.