Ayush Gupta

Built an AI-based object detection model to identify safety helmets in manufacturing environments, using deep learning and Explainable AI (XAI) to visualize model focus and decision rationale.

Used Explainable AI (XAI) to see that the model’s layer focused not only on helmets but also showed bias toward safety vests. Developing an Agentic AI safety system that autonomously monitors manufacturing video streams, detects helmet compliance using YOLOv8, applies Explainable AI (Grad-CAM) for decision transparency, and generates safety compliance reports.

Fine-tuned Large Language Models (LLMs) for Out-of-Distribution (OOD) detection using synthetic adversarial datasets, achieving AUC scores of 0.99 for near- and far-OOD tasks, significantly improving model reliability.

Developed a modular evaluation pipeline to identify and benchmark LLM failures under distributional shifts, contributing to LLM observability, safety, and real-world deployment readiness.

Built a personalized nutrition app with a chatbot by leveraging scientific nutritional research and FDA data to provide personalized food recommendations based on users' dietary preferences, health goals, and nutritional needs.

Improved context with OpenAI API, Google Maps APIs and Prompt Engineering.

Developed a stock price analysis application by using PySpark for Data Wrangling, Spark SQL for data querying, and ETL operations, and Delta Lake for reliable and scalable data storage and deployed using Kubernetes.