Mohamed Sabek
PhD Candidate at University of Alberta

AI-Driven Vision for Smarter Construction

Bridging computer vision, vision-language models, and lean construction principles to transform how we monitor, manage, and optimize construction sites — building systems that see, understand, and act in real time.

Get in Touch
17+ Publications
120K+ Dataset Images
7+ Years Experience
$30M+ Projects Managed

Live Monitoring

How VisioConTech Sees a Site

Drag through the day — the system tracks workers, equipment, and idle time frame by frame.

 VISIOCONTECH · LIVE06:00
  • DETECTING  WORKER · conf 0.98
  • FLAGGING  IDLE EXCAVATOR · 14 min
  • TRACKING  CRANE LOAD · conf 0.88
  • MEASURING  CYCLE TIME · zone B
  • ALERTING  CONGESTION · gate 3
$0Saved / site / yr
0%Safety score
0%Idle time

Fine-tuned YOLO detection fused with vision-language reasoning (Florence-2, Gemma 3) flags idle equipment, congestion, and worker activity in near real time — turning raw site video into lean, actionable metrics, all running on the edge.

Research Focus

Automated Vision for Lean Construction

Developing an integrated framework that synergizes computer vision and vision-language models to support data-driven lean construction management.

01

Lean Construction Visual Taxonomy

A validated, structured taxonomy mapping visual elements like idle equipment, congested pathways, and worker actions to fundamental lean principles and workflow indicators.

Lean Principles Taxonomy Expert Validation
02

CIVAD-v2 Dataset Expansion

Expanding the Construction Industry Vision Alberta Dataset to over 100,000 images and 100+ object classes, with rich annotation formats tailored for training advanced VLMs.

120K+ Images 100+ Classes VLM Captions
03

Integrated CV-VLM System

A software pipeline combining fine-tuned YOLO object detection with zero-shot VLM reasoning using Florence-2 and Gemma 3 for near real-time site activity identification.

YOLO Florence-2 Gemma 3 Edge Computing
04

Lean Decision Support System

An interactive dashboard visualizing CV-VLM insights with Grad-CAM explainability, feedback mechanisms, and digital twin integration for real-time lean planning.

Grad-CAM Digital Twin Power BI

Publications

Research Output

Peer-reviewed contributions across top venues in construction technology, AI, and robotics.

Journal Articles

2
2025

Bridging perception and decision-making: An integrated computer vision and fuzzy system for construction workspace management

Automation in Construction

2025

Immersive Safety: Revolutionizing Construction Training with Virtual Reality and Behavioral Insights

Lean Construction Journal (LCJ) — In Press

Conference Papers

15
2026

Automated Cycle Time Analysis in Prefabricated Construction Using Hybrid Vision-Language Models and Motion Heuristics

43rd ISARC, Singapore

2026

A Lean Construction Visual Taxonomy (LCVT): A Foundational Framework for Bridging the Semantic Gap in Computer Vision

34th IGLC, Singapore

2026

Zero-Shot Evaluation of SOTA Vision-Language Models for Detecting Construction-Related Activities

CSCE Annual Conference, Quebec City

2025

Automated sustainable evaluation for construction contracts using machine learning

CSCE & CRC Conference, Montreal

2025

Immersive safety: Revolutionizing construction training with virtual reality and behavioral insights

CSCE & CRC Conference, Montreal

2025

Digital twin-based construction fire hazard recognition training system

42nd ISARC, Montreal

2025

Construction Industry Vision Alberta Dataset (CIVAD)

42nd ISARC, Montreal

2024

Enhancing Construction Site Safety and Efficiency with YOLO v8-Based Computer Vision Model

CSCE Annual Conference, Niagara Falls

2024

Improving Tomorrow's Construction Sites Safety and Efficiency with YOLO v8

IEEE ICRA, Yokohama (Poster)

2024

Digital Twin-Based Integrated Decision Support System

32nd IGLC, Auckland

2022

Roboflow 100: A Rich, Multi-Domain Object Detection Benchmark

arXiv Preprint

2022

A Performance Study of Different Deep Learning Architectures for Detecting Construction Equipment on Sites

MSc Dissertation, Arizona State University

Projects

Built & Shipped

From large-scale datasets to AI-driven monitoring platforms — applied research turned into real systems.

Featured Mitacs Funded — $15K

VisioConTech

AI-driven real-time construction monitoring system incubated at the University of Alberta through Mitacs Lab2Market. Combines computer vision with lean construction metrics to provide actionable safety and productivity insights, with potential savings of $350K annually per site.

Computer Vision YOLO Edge Computing OpenVINO TensorRT Lean Construction
LIVE ·  Detecting worker · conf 0.98

CIVAD Dataset

One of the most extensive construction-specific computer vision datasets — 120,000+ annotated images across 100+ object classes from active Alberta construction sites.

120K+ Images Open Data

Digital Twin DSS

Digital Twin-based decision support system integrated with the Last Planner System for real-time constraint removal and resource planning on construction projects.

Digital Twin LPS

VR Safety Training

Immersive virtual reality system for construction safety training, incorporating behavioral insights and fire hazard recognition with VLM-driven scenario generation.

VR Safety VLM

Automated Grading System

LLM and computer vision-based automated grading for engineering mechanics assignments, enhancing assessment efficiency in undergraduate education.

LLM EdTech

Sustainable Contract Evaluation

Machine learning framework for automated sustainability evaluation of construction contracts, supporting greener procurement decisions in the construction industry.

ML Sustainability

Beyond Construction

One System, Many Venues

The same monitoring pipeline generalizes from construction sites to stadiums, warehouses, and beyond.

VISIOCONTECH · GENERALIZED
CONSTRUCTION SITE
12% idle equipment

Experience

Background & Journey

2023 — Present

Research Assistant

University of Alberta, Edmonton, Canada

Conducting advanced research on AI-based monitoring for construction. Leading the CIVAD dataset creation (120K+ images). Published in CSCE, IGLC, ISARC, and IEEE ICRA.

Teaching Assistant — 4 courses: Engineering Mechanics, Construction Methods, Mechanics of Deformable Bodies, Contract Administration (Head TA)

2016 — 2023

Project Manager & Civil Engineer

Wadi Serga Est., Saudi Arabia

Supervised 12+ infrastructure projects totaling $30M+. Led design, geotechnical investigations, and cost/risk management strategies across residential and commercial developments.

Education

PhD in Construction Management
University of Alberta — 2023–Present
MBA
University of South Wales, UK — 2022–2023
MSc in Construction Management
Arizona State University — GPA 4.0/4.0
BSc in Civil Engineering
Prince Sattam Bin Abdulaziz University

Awards

APIC Energy Hackathon — 1st & 2nd Prize (2025)
Alberta Innovates Scholarship — $31K/yr (2024–2027)
AWS Wildfire Hackathon — 1st Place (2024)
CIC ELEVATE Competition — 2nd Place (2024)
Al Ghurair Scholarship — $35K (2020–2022)
Pitchfork Changemaker Award — ASU (2024)

Certifications

PMP RMP CSCE Member ASCE Member

Leadership

Chief Faculty Steward, GSA Labour Relations
GSA Councillor-at-Large, University of Alberta
Board Member, OSPF Edmonton
Committee Member, CSCE Engineering Management

Technical Skills

Python YOLO Detectron2 TensorFlow PyTorch Florence-2 Gemma 3 OpenVINO TensorRT R SQL Power BI GIS Revit AutoCAD