Laxminarayana Yaga

Led development of supervised and unsupervised ML models in PySpark, TensorFlow, and XGBoost to detect fraudulent transactions, reducing false positives by 18% and improving detection. Optimized credit risk scoring with PCA and ensemble methods, boosting predictive performance by 12%. Built NLP pipelines using BERT, GPT-4, and NER to automate document classification and regulatory checks, cutting manual review time by 30%. Designed a Retrieval-Augmented Generation (RAG) system with LangChain, LLaMA, and LlamaIndex to process over 5 million records, increasing analyst research efficiency by 40%. Automated model evaluation and MLOps on Azure with CI/CD, reducing deployment cycles from weeks to days. Created real-time Power BI dashboards with PostgreSQL backend to monitor loan portfolio health and flag high-risk loans earlier.

Prototyped and evaluated multiple deep learning architectures using Keras and PyTorch, accelerating development cycles by 20%. Engineered feature engineering and time-series forecasting to improve demand accuracy by 14%. Built ML pipelines with Decision Trees, Random Forests, Naive Bayes, K-Means, DBSCAN, delivering client-specific predictive and clustering solutions that improved decision-making by up to 35%. Implemented CNNs and NLP workflows with BERT, NLTK, and Spacy, enhancing text classification and sentiment analysis on large-scale feedback datasets by 20-25%. Managed unstructured data in MongoDB and deployed production-ready models on AWS SageMaker, Lambda, and CloudFormation, achieving 99.9% availability. Created dashboards in Plotly and Tableau for real-time KPI monitoring, reducing reporting lag by 40%. Integrated FastAPI and Streamlit interfaces to operationalize ML apps, cutting manual intervention by 30%. Optimized data handling with Pandas, NumPy, SciPy, reducing pipelin