Bengazi Radar: Redundancy Risk Prediction System
A client-facing machine learning project for Bengazi Career Engagement. The work focused on preparing the dataset, training and comparing multiple candidate models, tuning hyperparameters and documenting the final prediction workflow with responsible AI considerations.
- Context
- Client-facing coursework project
- Focus
- Redundancy risk prediction
- ML work
- Model comparison · tuning · evaluation
- Deliverables
- Pipeline · architecture · responsible AI report
Overview
Bengazi Radar was built for Bengazi Career Engagement as part of my Responsible AI module. The project explored how a machine learning system could support redundancy risk analysis while keeping the modelling process explainable, documented and suitable for a client-facing decision support context.
The technical workflow covered dataset preparation, feature handling, model training, algorithm comparison, hyperparameter tuning and evaluation. The final approach was selected based on model performance, interpretability and suitability for the business problem rather than a single metric alone.
System concept
Step 01
Client problem
Redundancy risk context
Step 02
Data preparation
Cleaning · features
Step 03
Model training
Multiple ML algorithms
Step 04
Tuning
Hyperparameters
Step 05
Evaluation
Comparison · limitations
Step 06
Architecture
System design diagram
Step 07
Documentation
Responsible AI portfolio
What I built
- A predictive machine learning workflow for redundancy risk analysis.
- Training and testing across multiple candidate algorithms, including an XGBoost-based approach.
- Hyperparameter tuning to improve model performance and compare candidate configurations.
- An evaluation workflow for comparing models and communicating performance, trade-offs and limitations.
- A system architecture diagram explaining how the proposed prediction workflow would operate end to end.
- A Harvard-referenced responsible AI portfolio covering project framing, model choices and safeguards.
Why it matters
This project is useful portfolio evidence because it sits close to real business decision-making. The technical work was not only about fitting a model. It also required communicating how the system should be used, where it could fail and what safeguards matter when predictions may affect people.