Machine Learning Intern
Nexlyra Engineering Works · Aug 2026 – Present
- DevelopedAI-drivenrisk, frauddetection, propertyvaluation, andclassificationsolutionsbytranslatingbusinessrequirements
- into machine learning workflows and production-oriented models.
- • Built an end-to-end real-time fraud detection system using 500K+ transactions, engineering behavioral, temporal, and
- transaction-velocity features for risk scoring and fraud prediction.
- • Trained and evaluated XGBoost and LightGBM classification models, addressing severe class imbalance using SMOTE
- and optimizing model performance using cross-validation and hyperparameter tuning.
- • Applied SHAP Explainable AI techniques to interpret model predictions, identify high-impact features, and translate
- model behavior into business-oriented risk insights.
- • Developedregressionandclassificationmodelsusing 1M+propertyrecordsforautomatedpropertyvaluationandhigh-risk
- asset identification.
- • Implemented end-to-end Scikit-Learn pipelinescovering data preprocessing, missing-value imputation, feature engineering,
- categorical encoding, scaling, model inference, and deployment readiness.
- • Performed model evaluation using classification and regression metrics, analyzed model errors and feature importance, and
- iteratively improved model performance based on business requirements.
- • Collaborated with technical and business stakeholders to define ML use cases, analyze model performance, identify optimiza-
- tion opportunities, and align AI solutions with operational objectives.

