Praneeth Padala

Designed and implemented an enterprise-grade Generative AI Knowledge Assistant to support fraud, compliance, and risk teams by retrieving context-aware insights from policies, claims, and transactional document repositories. Built an end-to-end GenAI and agentic architecture spanning ETL pipelines, document embeddings, Retrieval-Augmented Generation (RAG) workflows, LangChain- and LangGraph-based orchestrations, MCP-style tool interfaces, backend APIs, and cloud deployments on AWS, with experiment tracking and governance using Azure ML and Vertex AI. Engineered scalable Python-based microservices to extract, normalize, and preprocess large volumes of regulatory, policy, and claims documents stored in Amazon S3, enabling downstream conversational AI fulfillment services. Processed unstructured data with OCR, tokenization-aware splitting, and metadata enrichment to improve embedding quality and grounding. Performed LoRA / PEFT fine-tuning on transformer models (Llama-3) to boost domain-s

Led the development of Clinical Insights and Readmission Prediction Platform leveraging ML, NLP, and DL to identify high-risk patients and optimize care management workflows. Built an end-to-end architecture for data ingestion, feature engineering, model training, inference, and analytics with strict healthcare compliance and governance. Ingested and harmonized multi-source data (EHRs, claims, unstructured clinical text) using Airflow, PySpark, and AWS Glue. Engineered NLP pipelines with BioBERT/BioGPT, spaCy, and transformers for NER and clinical text summarization. Modeled disease progression and readmission risk with LSTM, GRU, and Transformer architectures. Automated prior authorization/classification workflows. Built risk stratification and patient segmentation models with XGBoost, Random Forest, and logistic regression. Deployed Generative AI summarization tools (T5, GPT) for clinician-facing insights. Implemented robust MLOps with MLflow, Docker, and Kubernetes, including drift

Built AI-driven analytics for state-level claims processing, fraud detection, and operational decision-making across AWS and Snowflake environments. Ingested and harmonized large-scale data from state claims systems using REST APIs, AWS Glue, and Airflow; performed PySpark-based transformations; centralized curated datasets in Snowflake for scalable reporting and experimentation. Engineered predictive features for fraud indicators, utilization anomalies, and disease patterns; trained models with XGBoost, LightGBM, Random Forest, and Logistic Regression, with rigorous cross-validation and automated experimentation using MLflow and SageMaker. Deployed models as FastAPI REST services with automated retraining and monitoring; implemented security practices for data handling and governance. Built dashboards in Power BI and Grafana for real-time monitoring of model performance and operational KPIs.