Mounika Narra

Architected and deployed a scalable Retrieval-Augmented Generation (RAG) platform leveraging LLMs, LangChain, and vector search to generate context-aware financial insights. Built semantic retrieval pipelines with embeddings (FAISS/Pinecone), integrated knowledge graphs for enhanced contextual understanding, and implemented governance and evaluation workflows. Delivered scalable AI microservices via FastAPI, Docker, and Kubernetes for high-throughput production inference; established real-time and batch data ingestion with Apache Kafka; tuned prompts and caching to optimize latency and response quality; implemented MLflow-based experiment tracking and model versioning.

Designed and developed ML models for clinical risk prediction and claims intelligence using XGBoost and Scikit-learn. Built scalable data processing and feature engineering pipelines with Apache Spark/PySpark to handle large healthcare claims data. Deployed real-time risk scoring APIs using FastAPI, and implemented model monitoring, drift detection, and evaluation pipelines with MLflow and Evidently AI. Applied SHAP for explainability to support audit/compliance, and automated ML pipelines with Apache Airflow on AWS to streamline training, deployment, and lifecycle management.

Designed and developed machine learning models for demand forecasting and personalized product recommendations. Built scalable data processing and feature engineering pipelines (Spark, Pandas) and implemented forecasting models to support planning across product categories. Developed REST-based inference services using Flask and conducted model evaluation and A/B testing to improve relevance and business impact.