Amandeep – Python, LLM, NumPy

Built and productionized a scalable credit risk prediction system using machine learning and deep learning to enable real-time decision-making and automate risk assessment.

• Built baseline models (Logistic Regression, Random Forest, XGBoost) and benchmarked performance;
• Designed deep neural networks using TensorFlow and Keras to capture complex, non-linear relationships in financial data;
• Performed advanced feature engineering on transactional, behavioral, and credit history data;
• Handled class imbalance using SMOTE and cost-sensitive learning techniques;
• Developed end-to-end ML pipelines using MLflow for experiment tracking, model versioning, and reproducibility;
• Processed large-scale datasets using PySpark and Databricks;
• Deployed models on Azure/AWS with real-time scoring capabilities;
• Implemented model monitoring (drift detection, performance tracking) to ensure long-term reliability

Key achievements:

• Improved model accuracy from 67.4% to 82%;
• Reduced false negatives in credit risk detection (critical for financial loss prevention);
• Delivered ~18% YoY business impact through improved risk-based decisioning;
• Enabled a scalable, production-ready ML system used by risk and product teams.

Developed forecasting and classification models to predict client engagement and optimize operational resource allocation.

• Developed and optimized time-series forecasting models to support business planning;
• Designed and automated model retraining pipelines using Azure ML;
• Deployed machine learning models as scalable REST APIs for production use;
• Monitored model performance and detected drift to ensure reliability and accuracy;
• Improved operational planning and resource utilization through predictive insights;
• Enabled automated and scalable ML deployment across the organization.

Designed a scalable experimentation framework to evaluate product features and optimize acquisition strategies.

• Built and maintained an A/B testing framework, including hypothesis testing and power analysis;
• Defined KPIs and guardrail metrics to ensure reliable experiment evaluation;
• Performed uplift modeling and user segmentation to identify high-impact opportunities;
• Partnered closely with product teams to design and run experiments;
• Increased qualified leads by 20% through data-driven experimentation Enabled data-informed product decisions across key initiatives.

Designed and developed scalable data pipelines and dashboards enabling near real-time business monitoring.

• Built and maintained scalable ETL pipelines using Python, SQL, and Spark;
• Designed and delivered curated datasets to support analytics and machine learning use cases;
• Developed interactive Power BI dashboards with clear KPI tracking and visualization;
• Integrated data workflows with cloud storage solutions, including Azure Data Lake and AWS S3;
• Enabled near real-time visibility into key business metrics for faster decision-making;
• Significantly reduced manual reporting efforts through automation and data pipeline optimization.