Developed a guardian system for prompt injection prevention in LLM environments, featuring an API Gateway, a flexible processing pipeline within the guardian engine to other referee LLMs and a robust database architecture resumes and predicted the best candidature to make informed decisions.
Collected and preprocessed historical transaction data, conducted exploratory data analysis (EDA), and developed Ran- dom Forest and XGBoost with optimized hyperparameters to accurately detect fraudulent transactions. • Applied CI/CD with GitHub Actions, containerized the model using Docker, and deployed it on the Azure platform with a REST API for real-time processing. Set up monitoring tools for performance tracking and anomaly detection.
Developed a system using Retrieval-Augmented Generation and large language models to enhance document search, achieving a 90% accuracy in retrieving relevant sections by leveraging semantic similarity and context awareness. • Implemented RAG techniques to integrate a dense vector search with a pre-trained language model, enabling the system to return highly relevant document snippets rather than generic keyword matches.
A company wants to automate the loan eligibility process based on the information provided by the customer. To automate this process, we seek to identify customer segments, those who are eligible for the loan amount.
