I enjoy making things. Here are a selection of projects that I have worked on over the years.
Used OpenAI's ChatGPT-3.5 with Retrieval-Augmented Generation (RAG) on a collection of recent and personally selected AI safety papers.
A small project were I redteam differetn LLMs models with type of attacks.
Leveraging adversarial perturbations as diagnostic probes to enhance the performance of models on out-of-distribution (OOD) detection.
A high-level overview of 3D physical adversarial attacks designed to fool AI object detection systems in real-world conditions.
Different Active Learning strategies for selecting the most informative images to train object dectors models.
