Desmond L.

Key Skills

Software

Internal/Proprietary Tooling

OpenCV AI Kit (OAK)

Other

Top Subject Matter

Computer vision ROI detection for embedded/TV software testing

3D point cloud meshing and geometry reconstruction

Top Data Types

Image

Video

3D Sensor

Top Task Types

Classification

Tracking

Point/Key Point

Segmentation

Freelancer Overview

Embedded Systems Software Engineer (Sony EMCS) — AI + image processing ROI automation. Brings 4+ years of professional experience across complex professional workflows, research, and quality-focused execution. Core strengths include OpenCV AI Kit (OAK) and Other. Education includes Bachelor of Science (Honours), University of Nottingham Malaysia (2024) and Foundation in Engineering, University of Nottingham Malaysia (2021). AI-training focus includes data types such as Image and 3D Sensor and labeling workflows including Point, Key Point, and Segmentation.

Labeling Experience

Embedded Systems Software Engineer (Sony EMCS) — AI + image processing ROI automation

OpenCV AI Kit (OAK)ImagePoint/Key Point

Implemented AI + image processing workflows to identify ROIs and support debugging automation. Used these ROI detections to accelerate bug reproduction and reduce time spent diagnosing issues. Focused on converting visual signals into actionable outputs for engineering workflows. • ROI identification for image frames • Automated bug reproduction using AI outputs • Debugging acceleration through faster issue isolation • Integration of AI image processing into embedded software testing pipelines

2025 - Present

AI Data Annotator

Internal/Proprietary ToolingImageBounding Box

Worked with AI training data and data labeling tasks, focusing on maintaining accuracy and consistency across large datasets. Equipped with strong familiarity on common annotation tools and understand the importance of clean, well-structured data in training reliable AI models.

2025

Research on 3D Point Cloud Meshing — 3D data labeling via reconstructed meshes

Other3D SensorSegmentation

Developed an end-to-end pipeline to reconstruct 3D watertight meshes from 2D imagery using multiview stereo. Applied post-processing for noise reduction, outlier removal, rasterisation optimization, and curvature-aware smoothing to improve spatial accuracy. Evaluated meshing outputs with quantitative benchmarks and improved mesh completeness using neural implicit representations. • 2D-to-3D reconstruction for creating watertight meshes • Post-processing to refine 3D geometry • Benchmark evaluation (e.g., Hausdorff distance, vertex density) • Neural implicit modeling to improve mesh completeness

2021 - 2024

Education

Work History