Praneeth Karra

Designed and deployed ML models to predict fraudulent transactions and credit risk, improving detection accuracy by 21%. Developed data ingestion pipelines using Apache Airflow and AWS Glue to process large-scale financial datasets from multiple sources, improving data availability by 40% and reducing integration errors by 25%. Engineered data preprocessing workflows in Pandas and PySpark to enhance model training efficiency, reducing processing time by 30% and increasing pipeline throughput by 20%. Implemented NLP algorithms in SpaCy & Hugging Face Transformers to extract entities from financial documents, contracts, and reports, accelerating document analysis. Developed and trained deep learning models (CNN, RNN) for pattern recognition in transaction data, leveraging transfer learning to detect anomalies and reduce false positives. Deployed ML models as RESTful APIs using Flask & Docker on AWS SageMaker, enabling real-time fraud detection and credit scoring with sub-200ms latency an

Built predictive models using Python (scikit-learn) to identify customer churn and retention strategies, improving retention. Fine-tuned and optimized Large Language Models (LLMs) using prompt engineering and transfer learning to improve contextual accuracy and deliver domain-specific insights. Automated ETL workflows in Python and SQL, streamlining data preprocessing and reducing pipeline runtime by 35%. Conducted exploratory data analysis (EDA) and applied statistical techniques to uncover business insights from structured and unstructured datasets. Employed ML models (Random Forest, KNN, Naive Bayes) to optimize loan approval processes, reducing default rates by 20% and improving portfolio profitability. Applied NLP techniques (sentiment analysis) using Azure Cognitive Services to analyze customer feedback, driving improved service strategies. Optimized time-series forecasting models for demand prediction with good forecast accuracy. Created interactive dashboards in Tableau and Pow