Projects
Sep 2026 – Present
Paper2Video
Generates narrated explainer videos from research papers and lectures. A set of reusable primitives — extract, plan, narrate, render, deliver — turns a paper into a scripted, scene-by-scene video, applied both to papers I picked from the literature and to my own first- and second-author work.
Jan 2026 – Present
Jailbreaking Multimodal Unlearning Models
Investigating jailbreak and adversarial prompting strategies against multimodal unlearning models to evaluate unlearning reliability and identify failure modes under malicious or adaptive attacks.
Dec 2025 – Present
Multimodal Out-of-Distribution (OOD) Detection
Developing methods to detect out-of-distribution inputs across multiple data modalities, primarily focusing on Vision Language Models to improve AI robustness and reliability.
Sep 2025 – Dec 2025
Rationale-Augmented Knowledge Distillation for Small Language Models
Developed a multi-task knowledge distillation framework to train compact language models (e.g., T5-Small, FLAN-T5-Small) to jointly predict task labels and reconstruct LLM-generated rationales.
Dec 2024 – Present
Feature Selection in Manufacturing using Knockoff Framework
Implemented the knockoff framework for feature selection in manufacturing to select 44 relevant features out of 590 while controlling false discovery rates.
Apr 2024 – Jan 2025
Contrastive Learning Framework for Knowledge Distillation
Conducted robustness analysis against adversarial attacks and imbalanced data.