莉莎 | Rizal KA

Investigating how prompting strategies influence user productivity when interacting with AI tools like ChatGPT, Gemini, and DeepSeek.

• Conducted a comprehensive survey with 243 respondents to analyze usage patterns and prompt styles.
• Analyzed prompt structure influence on response quality and user workflow efficiency.
• Focused on real-world applications of LLMs in academic, creative, and professional environments.

A deep-learning pipeline that classifies food images using a Vision Transformer model and generates detailed recipes and nutritional feedback using the Gemini API.

• Built a high-accuracy food image classification system using EfficientNet-B4 backbone.
• Integrated Gemini LLM API to provide context-aware cooking instructions and nutritional guides.
• Developed full-stack architecture using a Python (FastAPI) backend and Vue.js frontend.

An analytical research project exploring the delicate trade-offs between imperceptibility and robustness in modern digital image watermarking algorithms.

• Investigated parameters such as PSNR (imperceptibility) and NC/BER (robustness) under spatial and frequency domains.
• Compared LSB (Least Significant Bit), DFT (Discrete Fourier Transform), and a custom LSB+DFT hybrid approach.
• Evaluated resilience against common attacks including JPEG compression, Gaussian noise, and salt-and-pepper noise.

A computer-vision and NLP-powered healthcare assistant designed to identify visible symptoms of common animal diseases and suggest treatment courses.

• Developed an image recognition model to accurately detect signs of visual pathology in animals.
• Integrated LLMs to process computer-vision outputs and generate actionable treatment reports.
• Created a complete end-to-end proof-of-concept pipeline demonstrating this diagnostic capability.