Ramya Sri Bobbilla

Developed and deployed generative AI applications using large language models, LangChain, OpenAI APIs, and RAG pipelines to improve enterprise knowledge retrieval accuracy by 37%. Built GPT-4-based knowledge assistant services using Pinecone, FAISS, and FastAPI, reducing support resolution time by 42% via automated document search and response generation. Fine-tuned open-source LLMs such as LLaMA 2 and Falcon using Hugging Face Transformers and PyTorch, improving domain-specific performance by 28%. Designed scalable AI service APIs with FastAPI and containerized deployment workflows using Docker and Kubernetes. Implemented MLOps lifecycle workflows using MLflow and Apache Airflow for experiment tracking, model validation, versioning, and deployment automation, along with prompt evaluation and response quality checks to strengthen production reliability.

Developed and deployed Generative AI applications using Large Language Models, LangChain, OpenAI APIs, and Retrieval-Augmented Generation (RAG) pipelines to improve enterprise knowledge retrieval accuracy by 37%. Built GPT-4-based knowledge assistant services using Pinecone, FAISS, and FastAPI, reducing support resolution time by 42% through automated document search and response generation. Fine-tuned open-source LLMs such as LLaMA 2 and Falcon using Hugging Face Transformers and PyTorch to improve domain-specific performance by 28%. Designed scalable AI service APIs with FastAPI and containerized workflows using Docker and Kubernetes for production serving. Implemented ML lifecycle workflows using MLflow and Apache Airflow for experiment tracking, model validation, version management, and deployment automation. Developed prompt evaluation and response quality checks to improve reliability and readiness for production.

Built machine learning solutions for demand forecasting and recommendation systems using Python, scikit-learn, XGBoost, and TensorFlow, improving prediction accuracy by 32%. Developed NLP pipelines with BERT, spaCy, and NLTK for sentiment analysis, text classification, and NER, improving F1-score by 27%. Created and integrated REST APIs using FastAPI and Flask to deploy models into production applications, reducing prediction latency by 38%. Developed feature engineering workflows to transform large-scale transactional and behavioral data into model-ready datasets. Optimized training and processing with Apache Spark, Apache Airflow, and Amazon SageMaker for data processing, training, hyperparameter tuning, and deployment.