AI ARCHITECT / ALGORITHM ENGINEER

Muhammad
Nabeel

I build computer vision, edge AI, and LLM systems for industrial inspection, robotics, and workflow automation.

Traffic monitoring
Traffic monitoring demo showing vehicle detections and activity charts
Computer vision dashboardExplore demos ↗

Project highlights

From data to deployment

The AI platform lifecycle, from camera datasets to edge devices.

  1. 01

    Collect & annotate

    Organize camera data into datasets and prepare labels for training and evaluation.

  2. 02

    Train & evaluate

    Train vision models, review metrics, and test predictions with held-out images or independent video.

  3. 03

    Develop & deploy

    Compose workflows, simulate models, and package compatible models for cloud or edge inference.

  4. 04

    Connect & monitor

    Connect device tasks and review detections, events, and operational reports.

Computer vision applications

Projects and application areas I contributed to as an AI Architect Engineer.

Smart factory

Factory cameras for process monitoring, visual inspection, and safety.

Assembly sequence anomaly detection (SOP monitoring)

Contributed to a camera-based pipeline that detects departures from expected assembly sequences.

YOLO · ONNX · Edge inference

Factory Kanban reporting

Built scheduled summaries of abnormal assembly sequences, with total counts and per-device breakdowns from algorithm results.

REST APIs · WeCom · Reporting

Fire and smoke detection

Contributed to real-time camera detection workflows for smoke and fire events.

YOLO · Video analytics

Pose and zone monitoring

Worked on pose-analysis and zone-exit workflows that turn camera events into alerts.

Pose estimation · Workflow orchestration

Industrial instrument monitoring

Contributed to the team’s computer vision work for capturing readings, recording data, and surfacing abnormal values.

Quality anomaly detection

Contributed to industrial visual inspection work for identifying quality issues in critical components.

Care and wellness

On-device perception and event-driven alerts for assistive scenarios.

Baby monitoring prototype

Tested on-device video analysis for scenarios such as a covered face or movement toward stairs.

View the proof of concept ↗

Horizon X5 · ROS2 · SmolVLM2

Assistive vision scenarios

Contributed to care and wellness projects, which includes obstacle awareness and reminders for health and exercise.

Assistive vision · Edge AI

Daily living assistance

Contributed to the team’s elder-care work for helping people track and manage everyday tasks at home.

Mobility safety monitoring

Contributed to computer vision scenarios for recognizing hazards in walking areas and generating alerts.

Independent living assessment

Participated in the elder-care application area for supporting assessment of instrumental activities of daily living (IADL).

Education and research

Dataset, training, and evaluation workflows that support experimentation.

Dataset and model management

Integrated dataset and model operations into the AI workflow platform, including catalog views and model metrics.

Datasets · Model metrics · REST APIs

Model simulation and deployment workflows

Integrated file-based model simulation, asynchronous result retrieval, and deployment plan operations.

Model evaluation · API orchestration

Industry customization

Tailored cloud-to-edge deployments, robot perception, and event automations.

ROS2 human-following robot

Led work on a perception and control pipeline that combines gesture recognition, detection, and tracking for autonomous following.

ROS2 · Robot perception · Tracking

RoboCar perception

Worked on camera-based perception and ROS2 integration for robotic applications.

ROS2 · Camera perception · Robotics

Multi-action gesture recognition

Contributed to a workflow that maps recognized gestures to distinct actions.

Computer vision · Gesture recognition

Camera-triggered automations

Developed reusable workflow configurations for printer completion, doorway loitering, and courier delivery.

Event detection · Workflow templates

Malan Lake wildlife detection

Worked on a 16-species detector, preparing YOLO datasets and evaluating recognition on separate wildlife videos.

YOLO · Dataset validation · Video evaluation

Edge inference optimization

Optimized and packaged lightweight computer vision pipelines for GPU and edge environments.

ONNX Runtime · TensorRT · Docker

Selected work

From training and evaluation to integration and deployment.

INDUSTRIAL COMPUTER VISION

Leather Defect Detection

Developed detection and segmentation pipelines for visual inspection of leather surfaces.

Read case study

Converted polygon annotations into masks, expanded and balanced the dataset, and compared YOLO and SAM-based approaches. Exported models to ONNX for the industrial inference pipeline.

View Leather-SAM results ↓

  • YOLO
  • SAM
  • ONNX
  • TensorRT

ROBOTICS & EDGE AI

ROS2 Human-Following Robot

Led development of a visual perception and control pipeline for autonomous human following.

Read case study

Combined gesture recognition, person detection, and visual tracking with robot control. The pipeline manages target selection and tracking as part of the following behavior.

Watch the RoboCar trials ↓

  • ROS2
  • Gesture recognition
  • Object tracking

MEDICAL IMAGING

Fundus Image Analysis

Evaluated detection and segmentation approaches for retinal imaging research.

Read case study

Converted ophthalmic annotations to VOC-XML, trained models, and inspected predictions through visualizations. Compared SAM-based methods, MMSegmentation, and YOLO across imaging tasks.

View Essex retinal imaging examples ↓

  • SAM
  • MMSegmentation
  • YOLO

Project demos

Retinal imaging, camera dashboards, RoboCar footage, and Leather-SAM results.

Computer vision dashboards

Traffic monitoring dashboard with car and bus detection boxes and traffic charts
Traffic monitoringVehicle detections and traffic activity in one dashboard.
Monitoring dashboard with a green box locating a detected wall crack
Wall crack detectionA camera feed with a localized wall crack and event counts.

RoboCar field tests

Indoor movement testRoboCar drives across an indoor floor beside the test laptop.
Indoor trial with a nearby objectRoboCar moves across the floor and comes to a stop near a bottle.

ESSEX BIOMEDICAL PHARMACEUTICAL

Diabetic eye imaging and retinal segmentation

I built and evaluated fundus image workflows for diabetic retinopathy research. The work included converting annotations, training and comparing models, and visualizing regions of interest such as microaneurysms, hemorrhages, exudates, and cotton wool spots.

PUBLISHED MEDICAL IMAGING RESEARCH

Dr.J breast ultrasound research

I co-authored the 2020 Dr.J paper on an AI-assisted preliminary screening system. The example shows how the research system outlined a finding in an ultrasound frame.

Published Dr.J ultrasound interface with a green outline showing a localized lesion
Lesion localization shown in the Dr.J publication.

INDUSTRIAL VISUAL INSPECTION

Leather-SAM

I adapted the Segment Anything Model for leather surface inspection by adding an attention mechanism to the encoder. I prepared LabelMe annotations, generated defect masks, compared predictions on leather samples, and exported the model to ONNX for future production use.

Intersection over Union
78.49%
Dice coefficient
84.91%

The project evaluation covered defects such as holes, cuts, scratches, folds, glue marks, and discoloration. Next steps documented in the project included expanding the labeled dataset and evaluating additional defect types.

Leather-SAM model architecture from the project slides
Model architecture
Examples of labeled leather defects with a color-coded legend
Annotated defect types
View 18 detection comparisons
Leather defect detection comparison 1Leather defect detection comparison 2Leather defect detection comparison 3Leather defect detection comparison 4Leather defect detection comparison 5Leather defect detection comparison 6Leather defect detection comparison 7Leather defect detection comparison 8Leather defect detection comparison 9Leather defect detection comparison 10Leather defect detection comparison 11Leather defect detection comparison 12Leather defect detection comparison 13Leather defect detection comparison 14Leather defect detection comparison 15Leather defect detection comparison 16Leather defect detection comparison 17Leather defect detection comparison 18

BACKGROUND

About me

I’m an AI architect and algorithm engineer with over six years of experience across Pakistan and China. I build computer vision, edge AI, and LLM systems, from training and evaluation through deployment.

Education

South China University of Technology
M.Eng. in Software Engineering · 2017–2019
COMSATS University Islamabad
B.S. in Software Engineering · 2013–2017

Technical skills

LLM systems
RAG, FAISS, tool calling, OpenClaw, n8n
Vision and learning
PyTorch, YOLO, SAM, OpenCV, reinforcement learning
Robotics and deployment
ROS2, SG2002, ONNX Runtime, TensorRT, Docker, Linux

Professional experience

Xeroptix Technology Limited, China
AI Architect Engineer · Oct 2025–Oct 2026
Designed LLM workflow orchestration, led ROS2 human-following development, and optimized edge inference.
Guangdong Ruizhou Information Technology
Algorithm Engineer · May 2024–Mar 2025
Built leather defect detection and segmentation pipelines and optimized model inference.
Essex Biomedical Pharmaceutical
Algorithm Engineer · Sep 2023–Mar 2024
Evaluated fundus image detection and segmentation models.
Shanghai Boonray / Daxing Hengchi Information Technology
Algorithm Engineer · Apr 2022–May 2023
Developed vision algorithms for vehicle detection and warehouse camera monitoring.
Fortex Solutions
Software Engineer · Jan 2021–Mar 2022
Led software delivery, mentored junior developers, and integrated AI detection modules.
iCare365 Technologies
AI Algorithm Engineer · Nov 2019–Oct 2020
Worked on model training and evaluation for preliminary breast cancer screening and image analysis.

MASTER’S THESIS · 2019

Cas-GAN for dialogue policy learning

I developed Cas-GAN to help task-oriented dialogue systems choose actions that complete user requests in fewer turns. The model combines graph convolution for dialogue representations, adversarial learning, and reward-based policy optimization.

South China University of Technology · Supervisor Prof. Wang Zhenyu

Task success rates reported in the thesis evaluation
Dialogue policyTask success
Cas-GAN
66.25%
Hierarchical reinforcement learning
64.85%
Reinforcement learning
44.40%
Rule-based agent
32.20%

Selected publications

CONTACT

Get in touch

I’m based in Pakistan and welcome inquiries about AI engineering roles, project collaborations, and applied AI research.

Send an email ↗