MLops Pipeline for Sentiment Analysis using LLMs

• Developed an end-to-end MLOps pipeline for sentiment analysis of Amazon customer reviews, ensuring scalable and automated workflows for data transformation, model training, and deployment.
• Fine-tuned large language models (DistilBERT, RoBERTa, ALBERT) for sentiment classification, achieving optimal performance based on accuracy, precision, F1-score, and other metrics.
• Automated data workflows using Apache Airflow for ingestion, preprocessing, and transformation, integrating data versioning with DVC and storage on Google Cloud Storage (GCS).
• Deployed the fine-tuned model on Vertex AI using CI/CD pipelines implemented with Jenkins for automated training, validation, and deployment.
• Designed monitoring and drift detection systems with Grafana, Google Cloud Monitoring, and Evidently AI to ensure model performance and reliability.
• Built an interactive Streamlit app hosted on Kubernetes for real-time sentiment analysis and visualization of customer insights and product performance.

• Github Project Link