Undergraduate Researcher at ALGOVERSE (2024-06 – 2025-01)
Conducted research on LLM prompting methods for sarcasm detection with published findings.
- First-author paper on "Pragmatic Metacognitive Prompting Improves LLM Performance on Sarcasm Detection" which was accepted to the 1st Workshop on Computational Humor (CHum 2025) at the 31st International Conference on Computational Linguistics. Presented work at international workshop.
- Developed prompting method that improved sarcasm detection performance of GPT-4o and LLaMA-3-8B by 10-12% over SoTA baselines across benchmark datasets.
- Surveyed 20+ papers on NLP sarcasm detection and LLM prompting; identified methodological gaps that guided experiment design.
- Implemented and ran hundreds of LLM evaluations using Python and Runpod, designing experiment pipelines to test prompt variants across multiple models and datasets.
National Science Foundation REU Intern at UNIVERSITY OF UTAH (2025-05 – Present)
Conducted research on medical image classification and segmentation with focus on model interpretability and performance optimization.
- Presented research poster findings at University of Utah summer symposium which diagnosed image classification and segmentation model limitations, identifying model behavior that did not focus on relevant pixels leading to poor generalization.
- Competed in MIDOG 2025 Challenge, applying advanced data augmentation, hyperparameter tuning, and cosine learning rate schedulers, achieving 80% accuracy for image classification models on distinguishing normal and atypical mitosis images for imbalanced and low resource datasets.
- Implemented and trained CNN-based medical image classification and segmentation models on MedMNIST using PyTorch, MONAI, and CHPC GPUs.
- Tracked experiments with CometML to systematically compare architectures and hyperparameters across 150+ runs.
Bakery Team Member at WHOLE FOODS (2024-10 – 2025-05)
Provided customer service and trained team members on bakery operations.
- Handled high-volume customer service and custom cake orders, consistently meeting deadlines and maintaining positive customer feedback.
- Trained 10+ temporary workers from other departments on bakery operations and product knowledge, ensuring consistent quality during peak periods.
