AI/ML Engineer & Data Scientist building business-facing machine learning systems for forecasting, fraud and risk scoring, churn prediction, operational intelligence, automation, dashboards, APIs, and decision support.
I build practical machine learning systems that go beyond notebook experiments: reproducible pipelines, time-aware validation, leakage-safe feature engineering, clear model evaluation, and deployment-oriented FastAPI or Streamlit interfaces where they add real value.
My work focuses on applied ML for business decisions: demand forecasting, fraud/risk scoring, churn prediction, anomaly detection, operational dashboards, and decision-support workflows. As a Project & Construction Management PhD student, I also bring domain context for construction risk and project analytics without limiting my work to that sector.
End-to-end fraud detection ML system focused on realistic validation, careful feature engineering, and deployable scoring components.
- Time-aware validation and leakage-safe feature engineering
- LightGBM, XGBoost, CatBoost, and weighted ensemble modeling
- Final ensemble ROC-AUC: 0.9292
- Includes FastAPI and Streamlit components
- Repository: peyamikenanoglu/ieee-fraud-detection
Retail demand forecasting system designed around time-based validation, business-relevant forecast quality, and an interactive demo.
- Lag features, rolling features, and time-aware validation
- LightGBM and CatBoost model comparison
- Champion LightGBM RMSLE: 0.373578 on internal time-based validation
- Live demo: retail-sales-forecasting-ml-system.streamlit.app
- Repository: peyamikenanoglu/retail-sales-forecasting-ml-system
Leakage-safe, time-aware churn modeling project emphasizing methodological correctness over inflated random-split scores.
- February-to-March validation setup
- Tuned XGBoost selected by LogLoss
- Focused on realistic customer churn evaluation and temporal generalization
- Repository: peyamikenanoglu/kkbox-churn-prediction
Secondary ML portfolio project covering movie revenue prediction and content-based recommendation.
- LightGBM, XGBoost, CatBoost, stacking, and recommender-system components
- Regression and recommendation workflows using movie metadata
- Repository: peyamikenanoglu/tmdb-box-office-ml-portfolio
Machine Learning: classification, regression, forecasting, churn prediction, fraud detection, anomaly/risk scoring
ML Engineering: reproducible pipelines, leakage-safe validation, feature engineering, model evaluation, FastAPI, Streamlit
Tools: Python, SQL, pandas, NumPy, scikit-learn, XGBoost, LightGBM, CatBoost, FastAPI, Streamlit
I am expanding this portfolio with a focused set of clean, real, portfolio-grade repositories built around practical AI/ML systems for business and technical review.
Current focus areas include forecasting, risk scoring, operational intelligence, document AI, RAG/LLM workflows, construction analytics, and client-facing AI systems.
- LinkedIn: linkedin.com/in/peyami-kenanoglu
- ORCID: orcid.org/0000-0003-1951-9534
- GitHub: github.com/peyamikenanoglu