Description: This project aims to preserve user privacy during interactions with large language models, which have a tendency to inadvertently leak sensitive information. Our current focus is on the Bangla language.

Description: This project explores intelligent information retrieval through question answering by leveraging knowledge graphs and graph neural networks. Our approach aims to challenge conventional models by incorporating structured semantic relationships and graph-based reasoning for improved accuracy and interpretability.

Description: Aimed to detect accident from video footage of a cctv-camera. I tried to develop alert system along with gps to inform authorities about the location of the accident. My model was able to detect accident from live camera as well as footage stored in the memory.

Description: Built a conventional CNN model to detect pneumonia accurately from chest X-rays. In addition, I also used a pre-trained model VGG-16 to classify whether the chest X-ray is normal or with pneumonia to compare the performance of CNN model.

Description: Designed for translating French to English. Here after going through several text preprocessing and cleaning I used Encoder-Decoder as the seq2seq model.

Description: Designed for an academic project where I used HTML, CSS, Javascript, ASP.NET and SQL Server for taking and maintaining the reservation of a Hotel.

Description: Designed a voltage supplier that provides 0-30v with over voltage, under voltage, short circuit protection capabilities and got introduced with different electronics components and PCB design.