Projects

A fine-tuned Stable Diffusion model that generates academic style figures from LaTeX captions, aiming to automate visual content creation for scientific papers.

A research project that used model-free reinforcement learning (Q-learning) to optimize household energy consumption based on grid load, electricity prices, and weather forecasts. Achieved 87% accuracy in demand prediction and reduced energy costs by 17%.

Implemented a custom C++ search engine using NLP techniques such as tokenization, stemming, stop-word removal, and TF-IDF weighting. Incorporated cosine similarity and basic synonym mapping to enhance result relevance. This project focused on efficient text preprocessing and scoring methods to deliver more accurate search results.