Usha Pavani Thopalle

Led the development of a GenAI-powered fleet operations assistant that summarizes telematics, GPS logs, and driver safety reports to automate risk scoring and compliance checks. Created fine-tuned large language models for summarizing extensive compliance policies, and implemented OpenAI function calling for enhanced bot interactions. Deployed scalable containerized AI services on Kubernetes and automated ML pipelines with MLflow, GitHub Actions, and Terraform. Developed vision models using CNN, transfer learning and self-supervised learning for defect detection. Mentored junior analysts on clustering and supervised machine learning techniques. Built classification pipelines leveraging XGBoost and SHAP for interpretable safety insights. Created data pipelines for large-scale IoT data processing and designed geospatial analytics models for risk zone identification. Introduced zero-shot and few-shot learning for adaptability on new scenarios, integrated monitoring tools for model drift a

Built a GenAI-powered fleet operations assistant leveraging LLMs to summarize telematics, GPS data, and safety reports for risk scoring and compliance. Developed automated compliance summarizers, integrated OpenAI Function Calling for enhanced decision support, and created internal copilots using LLM and RAG stacks for real-time guidance. Implemented monitoring frameworks with Databricks, MLflow, and Terraform to ensure reliable deployment at scale. Led multi-agent AI workflows using LangChain and Google ADK, refined image classification models, and fine-tuned LLMs for IoT datasets. Delivered predictive and classification models for driver safety and operational analytics. Conducted bias audits and enabled zero/few-shot learning for continual model adaptation. Mentored junior analysts and served as a customer-facing consultant deploying ML and RAG solutions.

Built a GenAI-powered fleet operations assistant integrating LLMs and vision models for telematics data summarization, compliance checks, and safety incident analysis. Developed automated compliance policy summarizers and internal GenAI copilots using RAG and LangChain frameworks. Deployed containerized ML models on Kubernetes clusters with CI/CD pipelines using MLflow, GitHub Actions, and Terraform on AWS. Mentored junior analysts on unsupervised learning and built explainable classification models to identify high-risk drivers and vehicles, reducing fleet incidents. Engineered large-scale IoT data pipelines in Databricks, performed feature selection and data augmentation to improve model recall, and designed bias detection audits to ensure fair AI recommendations in fleet safety. Integrated monitoring tools to track model drift and GPU usage for reliable production workloads.