Looking for a job

Ilyes Frigui

@ilyesfrigui

Message

Machine Learning Intern at BH Bank who improved fraud detection PR-AUC by 86% with a GIN and XGBoost ensemble.

Tunisia

Message

What I'm looking for

I'm seeking a final-year internship in Applied AI, Machine Learning, or Data Science.

At BH Bank, I compared XGBoost, GraphSAGE, GAT, and GIN for fraud detection across 5.1 million transactions. My GIN and XGBoost ensemble improved PR-AUC by 86% over the XGBoost baseline.

I also evaluated an unsupervised graph autoencoder, identifying 21 fraud cases ranked too low by the supervised model. I developed a technical decision framework comparing methods against data availability, compute resources, operational constraints, and business needs.

At Tunisie Telecom, I cleaned and analyzed telecom datasets with Python, Pandas, and NumPy to identify customer usage patterns and support KPI reporting. I implemented a KPI dashboard, explored forecasting for new SIM offers, and Dockerized the application.

In personal projects, I built an LLM-based data quality agent using profiling, semantic duplicate detection, clustering, and RAG over data dictionaries. I’ve also worked on IFRS 9 credit risk modeling and an LLM-enriched supply chain knowledge graph for route optimization.

Experience

Work history, roles, and key accomplishments

Machine Learning Intern

BH Bank

Aug 2026 - Sep 2026 (1 month)

Compared XGBoost, GraphSAGE, GAT, and GIN on 5.1M transactions (0.1% fraud), building a graph of 515k accounts with temporal validation to prevent future-edge leakage. Improved PR-AUC by 86% over the XGBoost baseline with a GIN + XGBoost ensemble, while analyzing when graph structure improved detection and when tabular features remained sufficient.

XGBoost GraphSAGE GAT Gin graph neural networks Fraud Detection Temporal Validation Ensemble Learning PR AUC Graph Autoencoders Python

Data Analysis Intern

Tunisie Telecom

Jun 2025 - Jul 2025 (1 month)

Cleaned, transformed, and analyzed telecom datasets using Python, Pandas, and NumPy, performing exploratory analysis to identify customer usage patterns and support KPI reporting for technical teams. Implemented a KPI dashboard and explored pre-launch forecasting of purchases and usage for new SIM offers, then Dockerized the application to ensure a reproducible and portable deployment environment.

Python Pandas NumPy Data Cleaning Exploratory Data Analysis KPI Dashboards Forecasting Docker