Thabitha Kaveti

Designed and deployed AI/ML pipelines on AWS SageMaker and Bedrock to automate healthcare claims validation, detect anomalies, and strengthen HIPAA-compliant workflows for large-scale claims processing. Built and optimized end-to-end ML pipelines (scikit-learn, TensorFlow, PyTorch, XGBoost) and data processing to detect fraud, segment providers, and classify claims. Developed Retrieval-Augmented Generation (RAG) architectures with LangChain, FAISS, and Neo4j to support contextual knowledge retrieval for provider and claims data. Fine-tuned Transformer LLMs (GPT, BERT, RoBERTa) on healthcare domain text, improving entity extraction and summarization. Implemented automated ETL with AWS Glue, Athena, and S3; orchestrated pipelines with Airflow, MLflow, and SageMaker Pipelines; containerized services with Docker and Kubernetes; built real-time inference APIs with FastAPI. Applied Explainable AI with SHAP and LIME; deployed model monitoring and compliance safeguards; built executive dashboa

Designed and deployed GenAI-powered claims automation pipelines on AWS SageMaker and AWS Bedrock, improving accuracy and reducing manual review time in large-scale healthcare transactions. Fine-tuned GPT, BERT, and Hugging Face Transformers for entity recognition, anomaly detection, and contextual knowledge retrieval. Built retrieval-augmented generation (RAG) pipelines with LangChain, FAISS, and LangGraph. Automated ML pipelines with Airflow, MLflow, and Kubeflow for reproducibility and monitoring. Deployed fraud detection models for real-time scoring, engineered big data pipelines on Apache Spark, and delivered scalable RESTful APIs and dashboards for leadership to monitor fraud risks and claims analytics. Ensured HIPAA-secure deployments with encryption and audit logging and applied SHAP/LIME explainability.

Designed AI/ML-based financial analytics on Azure. Built predictive pipelines in Azure ML and Azure Databricks for churn, asset-risk, and portfolio optimization. Developed time-series forecasting (ARIMA, Prophet, LSTM) in Azure Synapse. Finetuned FinBERT via Azure OpenAI for financial text summarization and sentiment. Used PySpark within Azure Databricks for feature engineering; performed Bayesian inference/Monte Carlo for portfolio risk and diversification. Leveraged Neo4j and Azure Cosmos DB for graph analytics; built deep learning architectures (TensorFlow, PyTorch) for trade classification and anomaly detection. Automated MLOps using Azure ML Pipelines, AKS, and DevOps; designed scalable data workflows with ADF/ADLS/Synapse. Implemented Bayesian optimization and built Power BI dashboards; incorporated Explainable AI. Explored Agentic AI and Prompt Flow for investment research summaries. Delivered a cloud-native AI/ML platform powering data-driven financial insights.