Lakshmi Mupparaju

Designed and deployed AI-driven chatbots, LLM-powered assistants, and predictive ML systems to enhance customer support and telecom operations. Implemented RAG-based AI chatbot with Pinecone; fine-tuned LLaMA and Mistral to reduce hallucinations; built real-time inference APIs (FastAPI & AWS Lambda); designed ETL pipelines (AWS EMR + Glue); OCR/NLP for invoicing; LangChain workflows; governance, vector stores, and retrieval-based retrieval; multi-agent GenAI workflows; and monitoring dashboards.

Designed and deployed AI-driven chatbots and LLM-powered assistants; implemented Retrieval-Augmented Generation (RAG) with Pinecone for telecom support; improved customer query accuracy by 40%. Fine-tuned LLaMA and Mistral on telecom data to reduce hallucinations and cut error rate by 25%. Built real-time inference APIs with FastAPI & AWS Lambda handling 100K+ queries per day with latency <200 ms. Implemented OCR+NLP pipelines for invoice/document parsing, saving 200+ hours per month. Created vector DB pipelines (FAISS, Pinecone, Chroma) for enhanced contextual retrieval. Migrated ML infrastructure to GCP (Vertex AI, BigQuery, GKE) reducing compute costs by 20%. Implemented MLOps with Kubeflow + MLflow and integrated RAG with knowledge-base retrieval. Mentored junior data scientists and collaborated with AI governance to onboard GenAI use cases.

Developed predictive ML models (Logistic Regression, Bayesian Regression, XGBoost) to optimize patient outcomes and reduce healthcare costs. Built NER pipelines with spaCy + AWS Comprehend for EHR entity extraction. Designed LLM-based summarization tools for clinical trial reports, reducing manual review time by 50%. Automated data pipelines with GCP Dataflow + PySpark, cutting ETL processing by 35%. Deployed multi-model serving pipelines on Vertex AI with automated rollback and access control. Migrated workloads to AWS SageMaker + Azure ML to improve latency and scalability. Built recommendation systems for personalized healthcare plans and implemented Kubeflow pipelines for MLOps automation. Reduced cloud costs via model compression and optimization. Collaborated with governance bodies to ensure HIPAA/regulatory alignment; on-boarded models into enterprise workflows.