Lakshmi Pulicharla

Designed and deployed AI/ML solutions for banking use cases including transaction monitoring, fraud investigation, and regulatory reporting using Python-based ML and deep learning models. Built, trained, and deployed ML models on AWS SageMaker, supporting batch and near real-time risk scoring across large-scale transactional datasets. Engineered end-to-end feature pipelines using AWS Glue, transforming raw transactional, customer, and behavioral data into model-ready features for training and inference. Containerized ML and inference services with Docker; provisioned cloud infrastructure with Terraform; integrated SageMaker inference endpoints using AWS Lambda for near real-time decisioning. Implemented CI/CD pipelines for ML workflows and applied statistical analysis, model evaluation, and threshold tuning. Established monitoring and logging for model execution and data drift to ensure reliability in regulated environments. Leveraged Generative AI capabilities via OpenAI GPT-4 for aut

Designed and developed machine learning and NLP models for healthcare use cases including patient risk stratification, readmission prediction, and clinical decision support. Built LSTM and CNN models to process structured data (vitals, lab results) and unstructured clinical notes. Developed AI-driven conversational and summarization solutions for EHR data using large language model frameworks, enabling clinicians to query patient records efficiently within secure environments. Prepared and curated healthcare datasets on Azure data platforms, performing data wrangling, normalization, and validation. Implemented NLP pipelines with spaCy and transformer-based models to extract medical concepts and insights from unstructured text. Applied dimensionality reduction and explainability techniques such as PCA and SHAP. Automated model training, evaluation, and deployment workflows using MLflow, with CI/CD on Azure Databricks. Orchestrated data processing and model retraining with Apache Airflow

Developed and deployed ML models for banking use cases including credit risk assessment, fraud detection, and customer analytics using Logistic Regression, Random Forest, and Gradient Boosting. Productionized models by converting research notebooks into scalable Python pipelines. Implemented feature engineering via PySpark transforming transactional and customer data into model-ready features. Built end-to-end ML pipelines across data prep, training, validation, and batch inference on enterprise cloud platforms. Fine-tuned models with cross-validation and regularization. Built RESTful inference services using Flask and integrated with CI/CD pipelines. Implemented monitoring for model execution, data drift, and logging to support regulatory readiness.