Marwan Ababneh
Artificial Intelligence Specialist
Management: Under a year
Python
SQL
JavaScript
Data Wrangling & Cleaning: 4/10
Predictive Modeling: 4/10
Active 6 days ago
Download CVCV
Overview
Technical skills
Projects
Timeline
Roles
Overview
A machine learning practitioner focused on end-to-end supervised modeling and simple dashboards at a middle level, with practical strength in implementing sklearn pipelines and Streamlit apps. The strongest proven skill is building and packaging predictive pipelines as evidenced by the ColumnTransformer + Pipeline pattern, stratified train/test split and joblib model export in train.py. The work lacks robust validation, tests, reproducibility tooling and explicit fairness or regulatory treatment for production loan decisioning.
Phone
Technical skills
Languages
3
Python
SQL
JavaScript
AI/ML
10
Pandas
Streamlit
Machine Learning
NumPy
Scikit-learn
PyTorch
TensorFlow
YOLO
Keras
CatBoost
Other
9
Flask
Matplotlib
Seaborn
GitHub
Git
Jupyter Notebook
Classic ML
Analytics
Computer Vision
Projects
Interactive Python Shop Application
Developed an interactive shopping application with a product catalog and a cart using Python. Implemented invoice saving functionality and organized application logic with object-oriented programming. Integrated product images into the user interface using Python image handling.
Smart Parking Management (YOLOv8)
Prepared the PKLot dataset for detecting available parking spaces in a university environment. Converted rotated bounding-box annotations into YOLO-compatible labels and created train/validation/test splits. Trained and tested YOLOv8n and YOLOv8m models using GPU support for computer-vision inference.
AI WebShield Extension
Co-developed an AI-assisted browser extension intended to improve web-surfing security. Worked with three team members to build the extension as an integrated solution for safer browsing. Focused on using AI-driven logic to support security-related behavior within the browser.
Timeline
Irbid National University
Bachelor's Degree ·
Artificial Intelligence and Data Science
Machine Learning Engineering Trainee
·
Junior
SDK Training Center
·
Full-Time
Built Python-based machine learning workflows covering data preprocessing, feature engineering, model training, validation, and evaluation. Investigated missing values and outliers and applied data transformations using Pandas and NumPy. Assessed predictive models with metrics such as MAE, RMSE, and R², and iteratively improved model performance. Developed interactive Streamlit apps to present data-driven results.
Python
Pandas
NumPy
Streamlit
Middle Data Scientist
Confidence: Medium ML Practitioner
A machine learning practitioner focused on end-to-end supervised modeling and simple dashboards at a middle level, with practical strength in implementing sklearn pipelines and Streamlit apps. The strongest proven skill is building and packaging predictive pipelines as evidenced by the ColumnTransformer + Pipeline pattern, stratified train/test split and joblib model export in train.py. The work lacks robust validation, tests, reproducibility tooling and explicit fairness or regulatory treatment for production loan decisioning.
Statistical Rigor
2/10
Correct use of statistics
Basic evaluation practice: stratified train/test split and accuracy reporting but no statistical testing, uncertainty quantification, calibration, or robust hypothesis checks.
Evidence
MarwanAb20/Loan_Approval/Model_Comparison.py: uses accuracy_score only for model comparison without confidence intervals or statistical tests
MarwanAb20/Loan_Approval/train.py: train_test_split with stratify but no CV or uncertainty estimates
MarwanAb20/Titanic-EDA-Project/EDA.py: EDA prints and plots without formal statistical checks
Data Wrangling & Cleaning
4/10
Preparing and cleaning data
Reasonable basic cleaning and preprocessing encoded as sklearn pipelines with imputation and scaling; limited provenance, leakage checks or advanced outlier handling.
Evidence
MarwanAb20/Loan_Approval/train.py: ColumnTransformer with numeric and categorical pipelines including SimpleImputer and StandardScaler
MarwanAb20/Loan_Approval/EDA.py: drops Loan_ID and reports missing values
MarwanAb20/Titanic-EDA-Project/EDA.py: fills Age with median, fills Embarked and drops Cabin
Exploratory Analysis & Visualization
3/10
Exploring and visualizing data
Contains useful visualizations and summary outputs but largely template-style plots with little written interpretation or data-driven storytelling.
Predictive Modeling
4/10
Building models that predict
End-to-end predictive workflow present with pipelines, multiple model comparisons and model export; lacks cross-validation, hyperparameter tuning and deeper error analysis.
Business Insight & Impact
2/10
Turning analysis into business value
Project is framed around a business problem (loan approval) but there is no evidence of cost-of-error analysis, fairness checks, regulatory considerations or actionable KPIs.
Reproducibility & Notebook Hygiene
2/10
Clean, repeatable analysis
Minimal reproducibility: random_state is used and model artifact is saved, but no environment pinning, tests, CI, data versioning or reproducible pipeline orchestration are present.
Expertise
Data Science· Middle
Industries
Financial Services· Middle
Technologies
Analytics
Classic ML
Python· since 2025 · Middle
Scikit-learn· since 2026
Seaborn
Matplotlib
Pandas· since 2025
Streamlit· since 2025
Machine Learning· since 2025
Recommendations
- Develop production-ready evaluation: add cross-validation, hyperparameter tuning, calibration, confidence intervals and more robust error analysis before deployment.
- Add reproducibility and engineering hygiene: pin dependency versions, include a requirements lockfile, add unit/integration tests and a simple CI workflow.
- Implement fairness and risk checks for financial models: bias testing, thresholding with FP/FN cost analysis and documentation for regulatory review.
- Convert Streamlit prototyping into a parameterized prediction API or pipeline for safer deployment and monitoring
Repositories
The developer's experience in this domain has been verified based on AI analysis of the following repositories:
Junior AI/ML Engineer
Confidence: Medium ML Engineer
A junior ML engineer focused on building end-to-end baseline classification pipelines and simple Streamlit dashboards; the single most important strength is assembling practical preprocessing and inference pipelines quickly. The strongest proven skill is implementing scikit-learn preprocessing and training pipelines and serving them with Streamlit, as shown in train.py and Prediction.py. What is not evidenced is robust production readiness, experiment tracking, automated testing, advanced model evaluation or custom model research.
Model Architecture & Training
3/10
How well models are designed and trained
Basic model-building and training using scikit-learn pipelines and sensible defaults but no advanced architecture, hyperparameter search, or robust validation.
Evidence
MarwanAb20/Loan_Approval/train.py: pipeline = Pipeline([('preprocessor', preprocessor), ('model', model)])
MarwanAb20/Loan_Approval/Model_Comparison.py: models dict with LogisticRegression, RandomForestClassifier, SVC, KNeighborsClassifier
MarwanAb20/Loan_Approval/train.py: RandomForestClassifier(n_estimators=300, random_state=42)
Data Pipeline & Feature Engineering
3/10
How data is prepared for models
Standard data preprocessing pipelines (imputation, scaling, one-hot encoding) and basic EDA; no advanced augmentation or leakage controls.
Evidence
MarwanAb20/Loan_Approval/train.py: ColumnTransformer with num_pipe (SimpleImputer, StandardScaler) and cat_pipe (SimpleImputer, OneHotEncoder)
MarwanAb20/Loan_Approval/EDA.py: missing value reporting and correlation heatmap
MarwanAb20/Titanic-EDA-Project/EDA.py: missing value handling and basic encoding steps
Experimentation & Evaluation
2/10
How results are measured and tested
Simple experimentation with train/test split, stratification and accuracy evaluation; no CV, no experiment tracking or reproducible experiment management.
MLOps & Deployment
2/10
How models are shipped to production
Minimal model lifecycle work: model persisted with joblib and a Streamlit app for inference; lacks versioning, CI/CD, monitoring or deployment hardening.
Computational Efficiency
1/10
How efficiently computing resources are used
No evidence of computational efficiency work such as GPU/parallel training, profiling, batching or quantization; only standard sklearn settings.
Research Depth & Innovation
1/10
Depth of research and new ideas
No research-level contributions, novel architectures, or paper reproductions; work is practical and tutorial-style.
Industries
Financial Services· Middle
Technologies
SQL
CatBoost· since 2026
Jupyter Notebook
YOLO
Computer Vision· since 2026
TensorFlow
NumPy· since 2025
Keras
Git
PyTorch
GitHub
Recommendations
- Develop production-ready ML services by adding model versioning, basic CI/CD and a small integration test suite for the training and prediction flow.
- Expand evaluation rigour: add cross-validation, holdout/validation splits, confusion matrices, and simple bias/fairness checks for the loan model.
- Practice experiment tracking and reproducibility using a lightweight tool (for example MLflow or simple structured logging) and store train/test metrics and model artifact metadata.
Repositories
The developer's experience in this domain has been verified based on AI analysis of the following repositories: