00:00:00
Available for opportunity

Abdullah Bakr, Computer Vision Engineer

Hi, I'm Abdullah Bakr. I build vision models that read medical images.

Open to internships, junior roles and freelance work, remote or hybrid.

Monufiya, EG · 2026

Core focus

Computer vision & medical imaging engineer

"Medical imaging. Computer vision.
models you can check."

I train models that find what matters in an X-ray or a skin photo, and I show where they look so a person can verify them.

Scope

Detection, segmentation and classification on chest and dental X-rays, breast ultrasound and skin images, from data cleaning to a working app.

From notebooks to usable tools.

How I work

Patient-level and lesion-level splits with no leakage, metrics on held-out data, and Grad-CAM checks on what the model is looking at.

Abdullah

Professional statistics

Images that speak. Models that explain. Code that ships. One engineer and the numbers behind the work.

Mean AUC
0.8495
PyPI downloads
17,500+
Public projects
17
Kaggle
Expert notebooks
About me
Scroll to exploreMy short story

My short story

Chest X-ray from the NIH ChestX-ray14 test set NIH ChestX-ray14 · test image Swin V2-B · Grad-CAM++ Truth: effusion attention 0%
Scroll to fade in where my chest X-ray model looks when it predicts effusion.

Building vision models
where medicine meets machine learning.

I'm a second-year Information Systems student in Egypt and a machine learning engineer focused on computer vision. Most of my work is on medical images: chest and dental X-rays, breast ultrasound and skin lesions. I spend a lot of time on the parts that make a number believable, then I turn working models into things people can use, like a dental X-ray app or the deepcsv library.

Tech stack & ecosystem

Feature — 01

MEDICAL
IMAGING

Chest and dental X-rays, breast ultrasound and skin lesions. Multi-label classification across 14 thoracic diseases, tooth-level dental analysis, and attention maps that show what the model used.

  • Grad-CAM++
  • Patient-level splits
  • NIH ChestX-ray14
  • DENTEX 2023
  • Focal loss
  • Multi-label

Feature — 02

DETECTION &
SEGMENTATION

Multi-stage pipelines where each model hands off to the next: YOLO finds the region, a Swin or ResNet classifier makes the call. Built for dental X-rays, skin lesions and drones in the wild.

  • Ultralytics YOLO
  • Segmentation
  • Swin Transformer
  • Two-stage pipelines
  • Albumentations
  • Test-time augmentation

Feature — 03

RESEARCH &
OPEN SOURCE

A first-author preprint benchmarking six architectures, an open-source library with 17,500+ downloads, public Kaggle notebooks, and 5+ ML challenges I designed and published on Shipd.

  • Benchmarking
  • Research writing
  • PyPI packaging
  • Kaggle Expert
  • Challenge design
  • Reproducibility

Discover my latest work and the models behind it

Scroll to explore

    More projects

    Professional Journey

    A timeline of my roles, research and study so far.

    B.Sc. Information Systems

    Sep 2025 – Jun 2029

    Science Valley Academy

    Second year, grade Very Good. Software development, databases and systems design, alongside my ML work.

    Research collaborator

    2026

    Chest X-ray study with Dr. Ashraf Ullah

    First author and sole implementer of a Swin Transformer V2-B study on NIH ChestX-ray14, now a preprint on Research Square.

    Read the preprint

    ML engineering intern

    Jul – Sep 2026

    FlyRank AI, remote

    Selected from 20,000+ applicants. Finished the full assignment track and a capstone on Google search ranking and discoverability, accepted by the lead mentor.

    Internship workspace

    ML engineer, freelance

    2026 – now

    Shipd, remote

    Solved 50+ ML challenges end to end. Designed and published 5+ original challenges, each with a deterministic dataset pipeline and reproducible public and private splits.

    Open source & community

    2026

    deepcsv, Kaggle, LinkedIn

    Maintainer of deepcsv (17,500+ downloads), Kaggle Notebooks Expert, and #8 in Egypt among AI and ML creators on Favikon (June 2026), writing for 14K+ followers.

    What people say

    He is a hardworking and consistent professional. We collaborated on a research paper together, and it was a pleasure working with him. He completed his part of the work within the given deadline. Abdullah has excellent programming skills in both Machine Learning and Deep Learning.

    Dr. Ashraf UllahResearch collaborator

    We recently collaborated on a Kaggle competition where we joined in the final 10 days, and his dedication was key to our success. … He took the lead on building robust ML pipelines and fine-tuning hyperparameter spaces using Optuna.

    Hozaifa MohamedCompetition teammate

    Certifications & achievements

    Learning that goes on the record.

    Courses and programs in deep learning, computer vision and cloud AI.

    Achievements page

    GitHub ecosystem

    Vision work,
    now open source.

    Every project on this page has public code: notebooks, training scripts and READMEs with the real numbers.

    Public repos
    25
    Followers
    13
    Since
    2025

    Kaggle

    Notebooks
    people actually read.

    Public notebooks for my medical imaging, detection and classification projects, each one runnable end to end.

    Tier
    Expert
    Category
    Notebooks
    Linked here
    12

    PyPI profile · AbduBakr

    Packages people
    pip install.

    Publishing on PyPI since March 2026, with 26 releases so far, from 0.1.0 to the current version.

    Packages
    1
    Total downloads
    17.5K+
    Releases
    26
    Latest
    0.8.0

    TRAININGmodels that see.

    More Projects

    Studies · Kaggle notebooks · Architecture comparisons

    Browse all repositories

      Let's Build Something
      Reliable Together

      Open to internships, junior roles and freelance work in computer vision and medical imaging.
      Send me the problem, I'll send back a plan.

      abdullah.bakr.official@gmail.com

      Certificate

      All certificates

      Send a message