Machine Learning Intern

Role Overview: You will be involved in designing and training fraud detection models for various types of fraud including application fraud, synthetic identities, mule accounts, and behavioral anomalies. Additionally, you will work on building credit risk and early default prediction models, handling highly imbalanced datasets, advanced feature engineering, model validation, monitoring, and continuous improvement. Your focus will be on core algorithm development, not operations or manual underwriting. Key Responsibilities: - Designing and training fraud detection models for application fraud, synthetic identities, mule accounts, and behavioral anomalies - Building credit risk and early default prediction models - Handling highly imbalanced datasets with precision-focused optimization - Performing advanced feature engineering from real-world loan and transaction data - Conducting model validation, monitoring, and continuous improvement - Collaborating with engineering teams to move models toward production readiness Qualifications Required: - Strong foundation in Probability, Statistics, and Linear Algebra - Clear understanding of ML fundamentals including bias-variance tradeoff, regularization, and cross-validation - Hands-on experience with classification models such as logistic regression, tree-based models, and gradient boosting - Strong Python skills - Hands-on experience with ML frameworks such as scikit-learn, TensorFlow, or PyTorch - Ability to work with large, messy real-world datasets - Independent thinking and ability to solve open-ended risk problems Additional Company Details: You will work on real NBFC data within a proper compliance structure. The models you build will directly impact approvals, capital allocation, and fraud losses. If you are interested in building risk engines that have a tangible impact on preventing fraud at scale, please send a DM with your CV and relevant project work. (Note: The contact information provided at the end of the job description has been omitted for privacy and security reasons) Role Overview: You will be involved in designing and training fraud detection models for various types of fraud including application fraud, synthetic identities, mule accounts, and behavioral anomalies. Additionally, you will work on building credit risk and early default prediction models, handling highly imbalanced datasets, advanced feature engineering, model validation, monitoring, and continuous improvement. Your focus will be on core algorithm development, not operations or manual underwriting. Key Responsibilities: - Designing and training fraud detection models for application fraud, synthetic identities, mule accounts, and behavioral anomalies - Building credit risk and early default prediction models - Handling highly imbalanced datasets with precision-focused optimization - Performing advanced feature engineering from real-world loan and transaction data - Conducting model validation, monitoring, and continuous improvement - Collaborating with engineering teams to move models toward production readiness Qualifications Required: - Strong foundation in Probability, Statistics, and Linear Algebra - Clear understanding of ML fundamentals including bias-variance tradeoff, regularization, and cross-validation - Hands-on experience with classification models such as logistic regression, tree-based models, and gradient boosting - Strong Python skills - Hands-on experience with ML frameworks such as scikit-learn, TensorFlow, or PyTorch - Ability to work with large, messy real-world datasets - Independent thinking and ability to solve open-ended risk problems Additional Company Details: You will work on real NBFC data within a proper compliance structure. The models you build will directly impact approvals, capital allocation, and fraud losses. If you are interested in building risk engines that have a tangible impact on preventing fraud at scale, please send a DM with your CV and relevant project work. (Note: The contact information provided at the end of the job description has been omitted for privacy and security reasons)