Open to opportunities

Abdullah Yousaf

@abdullahyousaf3

Message

Computer vision developer at Bentley Systems who built a GCP detector scoring F1 ≈ 0.92 on about 12,000 drone images.

Pakistan

Message

At Bentley Systems, I built a geometry-based ground control point detector that scored F1 ≈ 0.92 on about 12,000 drone images, without training data. I also built a PaddleOCR pipeline that turns equipment nameplate photos into structured JSON.

Before that, I labelled 3D point clouds and telecom tower images for Pointivo's ML team. At RISE Lab, NUST, I’m working on an MS thesis about reconstructing wheat plants in 3D and measuring their traits.

Experience

Work history, roles, and key accomplishments

Current

Consultant

Current

Bentley Systems

Dec 2025 - Present (10 months)

Built a ground control point detector for drone survey photos using geometry checks only, achieving F1 ≈ 0.92 on about 12,000 images. Developed an OCR pipeline for equipment nameplate photos using PaddleOCR, extracting structured data to JSON.

Python OpenCV NumPy Paddle OCR CVAT Azure Blob Storage Git Docker

Current

Graduate Researcher

Current

RISE Lab, NUST

Sep 2023 - Present (3 years 1 month)

Conducted research and course projects in computer vision and robotics, including an MS thesis on 3D reconstruction of wheat plants.

Computer Vision Robotics 3D Reconstruction Research

Data Annotation Specialist

Pointivo Inc.

Jun 2024 - Dec 2025 (1 year 6 months)

Labelled drone-captured 3D point clouds and photos of telecom towers, fitting 3D cuboids to antennas and other components. Trained new hires and drew AutoCAD site plans from top-view imagery.

3D Annotation Point Clouds Autocad Data Labeling Training