Vishnu Priya P

Master's

Computer Science

codelanceit

data scientist

 Built LightGBM-based churn prediction models with SHAP explainability, integrated with CRM to drive retention campaigns.  Developed BERT-based NLP pipelines to extract intent, sentiment, and key entities from unstructured support logs.  Deployed Prophet and ARIMA time-series models on AWS Lambda for forecasting key KPIs like default rates.  Orchestrated model training and deployment pipelines with GitHub Actions, MLflow, and Docker containers.  Built a distributed ETL pipeline using PySpark on Databricks, processing 1TB+ datasets daily from financial systems.  Created Streamlit apps for internal teams to interact with models and visualize scenario-based predictions.  Led containerization of AI services with Docker, deployed and monitored on Kubernetes clusters.  Implemented model monitoring dashboards using Prometheus + Grafana for drift and accuracy alerts.  Integrated model predictions into Power BI dashboards using Python and REST APIs for real-time decision-making.  Used TensorFlow with mixed precision and data augmentation to improve training times for fraud detection use case.  Designed and conducted A/B experiments to evaluate the business impact of ML-driven personalization strategies.  Collaborated cross-functionally with data engineers and business leaders to align technical solutions with use cases.  Conducted model retraining automation with DVC versioning and metadata tracking for audit trails.  Delivered quarterly stakeholder demos, showcasing model impact on KPIs like NPS, conversion, and CLTV.